Today, MarkTechPost had the pleasure of interviewing Joey Conway from NVIDIA to discuss their exciting work on open-source large language models, including Llama Nemotron Ultra & Parakeet. Highlights from the interview: NVIDIA’s Open Source Powerhouse: Discover how NVIDIA is pushing the boundaries of open-source AI with the release of cutting-edge models like Llama Nemotron Ultra and Parakeet TDT. Llama Nemotron Ultra: Smaller Size, Giant Performance: Learn how NVIDIA achieved on-par performance with models twice the size, enabling deployment on a single GPU node. Explore their innovative FFN fusion technique for significant speedups. Reasoning on Demand: Uncover the unique “reasoning on/off” feature in Llama Nemotron Ultra, offering unprecedented control for production deployments and cost optimization. Revolutionary Speech Recognition with Parakeet TDT: Dive into NVIDIA’s state-of-the-art ASR model that transcribes one hour of audio in one second with only a 6% word error rate – 50 times faster than other open-source alternatives! The “How”: Architectural Innovations: Get insights into the advanced architectures and optimizations behind these models, including FFN fusion, limited context attention, and the Token Duration Transducer (TDT)  Democratizing AI with Open Data: Learn about NVIDIA’s commitment to the open-source community through the release of model weights and massive, high-quality datasets for both language and speech. Future Directions: Get a sneak peek into NVIDIA’s plans for multilingual support, even smaller edge-optimized models, and advancements in real-time streaming for speech recognition. Production-Ready AI: Understand how these models are designed with real-world deployment challenges in mind, focusing on accuracy, efficiency, and cost-effectiveness. Jean-Marc Mommessin: Joey, welcome to Marketechpost! We’re thrilled to have you here and to delve into the impressive open-source models NVIDIA has been releasing. To start, could you please introduce yourself and your role at NVIDIA? Joey Conway: Hi Jean-Marc, it’s great to be here. I’m Joey Conway, and I work in product management for some of the deep learning software at NVIDIA. Our team focuses on large language models like Nemotron and Llama Nemotron, as well as text-to-speech models such as Parakeet. Jean-Marc Mommessin: Wonderful. And you’ve been at NVIDIA for over seven years now, witnessing significant waves of innovation in AI. Let’s talk about your recent release, Llama Nemotron Ultra, a 253 billion parameter model. From what we’ve seen, it delivers performance on par with models like Llama 405B and DeepSeek R1, which are about twice its size. Remarkably, it can run on a single 8x H100 node. What else can you tell us about Llama Nemotron Ultra and what makes it so impressive? Joey Conway: We’re big believers in the open-source community and the fantastic work being done there. With Llama Nemotron, our goal was to build upon the existing foundations, particularly Llama, for which we greatly appreciate Meta’s contributions. We also observed significant progress in reasoning within the open community earlier this year. Inspired by this, we wanted to contribute and see how we could enhance Llama, especially for enterprise use cases. Our focus was primarily on improving reasoning capabilities and agentic tasks like tool calling and chat. We aimed to take the strengths of the open-source community, enhance them, and then contribute those improvements back. Jean-Marc Mommessin: Did you identify specific gaps in existing models that you aimed to address? You mentioned reasoning, but could you provide an example or two of enterprise agentic tasks where you felt there were shortcomings that Llama Nemotron Ultra overcomes? Joey Conway : Yes, I think looking back to the beginning of the year, a key challenge in enterprise deployments was handling complex queries requiring significant thought and reflection. These could be multi-step processes or involve substantial calculations and the use of external tools. At that time, there weren’t many strong open-weight models capable of robust reasoning. The progress we’ve seen in the last few months in this area is very encouraging. Another critical aspect for enterprises is the ability to accurately call APIs and closely follow instructions in user queries. We wanted to ensure that while we focused on improving reasoning, we didn’t compromise these essential production-level capabilities. Furthermore, we often noticed that when both reasoning and instruction following were well-addressed, they typically resided in separate models. Our aim was to simplify this by creating a single model that excels in both. This was the landscape we observed when we started this project around January and February. Jean-Marc Mommessin: That makes perfect sense and aligns with what we’re seeing in the industry as well. Now, let’s dive into the “how.” Your paper mentions FFN fusion as a key optimization. Could you elaborate on this technique, starting with a high-level explanation? Joey Conway: Absolutely. Our focus on optimization stemmed from the realization that deploying state-of-the-art models often requires a significant deployment footprint. We wanted to optimize this to fit within more common GPU setups. We explored various techniques, including our Puzzle neural architecture search. For dense transformer models, particularly those in the Llama family, we discovered a way to reduce or eliminate redundant attention layers. This process aligned the feed-forward network (FFN) layers in a sequence, allowing us to explore fusion methods. Our fundamental goal on the GPU is to maximize parallel execution. Fusing these aligned FFN layers enables greater parallel computation than was previously possible. By removing redundant layers, we found opportunities to essentially merge or fuse the remaining ones. This is a key example of how we tackle the challenges of running these models at scale. Importantly, this technique often yields greater improvements with larger models, which was beneficial for our Ultra model based on Meta’s Llama 3.1 -405B. Jean-Marc Mommessin: And this FFN fusion significantly improves the model’s throughput, achieving notable speedups. If I recall correctly, it’s in the range of 3 to 5x for the Ultra model? Joey Conway: That’s right, the speedups for the Ultra model are in that range. Additionally, by reducing the model’s size in terms of weights, we also lowered its memory footprint. This allowed us to utilize a larger KV cache. For Llama Nemotron Ultra, we could fit it onto a 8x H100 80GB setup, which is quite significant as it fits within common node configurations. So, FFN fusion provided both a substantial compute speedup and a reduction in memory usage, enabling us to handle larger context lengths. These are very exciting outcomes for us. Jean-Marc Mommessin: Let’s switch gears to data curation. AI data is crucial, and your training pipeline seems very sophisticated. You touched on “instruction following” earlier. Could you elaborate on your data curation process and how you ensured high-quality data, especially considering you leveraged other models in the process? Image source: NVIDIA Joey Conway: Transparency and openness were key in our approach. We wanted to share as much as possible about our data, techniques, and tooling so the community could understand and even use it themselves. Our primary goal with data curation was to improve accuracy across several key domains, including reasoning tasks like math and coding, as well as non-reasoning tasks like tool calling, instruction following, and chat. Our strategy involved curating specific datasets to enhance performance in these areas. Within our supervised fine-tuning process, we differentiated between “reasoning on” and “reasoning off” scenarios. For example, in math and coding, we curated data for simple questions that don’t require complex reasoning, as well as more intricate problems that do. This helps the model learn when and how to apply reasoning. A key part of this process was leveraging high-quality models from the community as “experts” in specific domains. For instance, we used DeepSeek R-1 extensively for reasoning-intensive math and coding tasks. For non-reasoning tasks like basic math, coding, chat, and tool calling, we utilized models like Llama and Qwen. Our aim was to blend the best capabilities of these community models into a single model. We’ve also made this curated dataset publicly available on Hugging Face, with around 30 million question-answer pairs. This allows the community to explore, use, and build upon our work. We were also excited to see our partner ServiceNow recently announce their apprehend Nemotron model, which was trained using our dataset to enhance their own reasoning capabilities. Jean-Marc Mommessin: That’s fantastic that you’re sharing the dataset. Given that you used other models to generate some of this data, what kind of quality checks did you implement to ensure the reliability of the training pairs? Joey Conway: Data quality was absolutely paramount. Since we were generating a significant portion of the data using other models, we implemented a rigorous multi-layered quality assurance process. First, for each expert model used to generate data in a specific domain, we would generate multiple candidate responses for the same prompt. Then, we employed a separate set of “critic” models to evaluate these candidates based on correctness, coherence, and adherence to the prompt. Second, we implemented a scoring mechanism. Each generated question-answer pair received a quality score based on the critic model’s evaluation. We set a high threshold, and any pair that didn’t meet this standard was discarded. Third, human review was integrated at various stages. Our team of data scientists and engineers manually inspected samples of the generated data to identify any systematic errors, biases, or instances of hallucination. This human oversight was crucial for catching nuances that automated systems might miss. Fourth, we focused on the diversity of the generated data. We wanted to ensure we weren’t just getting variations of the same types of questions and answers. We implemented strategies to encourage the expert models to generate a broad range of examples within each domain. Finally, after training Llama Nemotron Ultra on this curated data, we conducted extensive evaluations against benchmark datasets and in real-world use cases. This feedback loop helped us further refine our data generation and filtering techniques. So, it was a comprehensive approach involving expert generation, automated criticism and scoring, human review, diversity checks, and rigorous downstream evaluation to ensure the high quality of our training data. Jean-Marc Mommessin: The quality of the synthetic data is so important. Could you elaborate on the stages you take to ensure high accuracy when generating this data? Joey Conway: Absolutely. When doing synthetic data generation, there are a few key stages to ensure high accuracy. The first is the prompts – the seed data and how we prompt the model. The second is the quality of the responses. On the prompting side, we focus on prompting models where we believe they excel. For example, we might use Llama for chat-related prompts but avoid using a non-reasoning model for math. It’s crucial to align the prompts with the core strengths of the model. For vetting the responses, we invest time in both human manual review and automated methods. Going forward, we anticipate increasing our use of verifiers and reward models, similar to what we’ve done on the Reinforcement Learning (RL) side. The reason we’ve open-sourced much of this is that there’s a lot of nuance involved, and we wanted the community to engage with these challenges. Enterprises like ServiceNow have specific goals, and some of our data might be more or less useful to them. By making it available, they can vet it themselves. We also provide tools like classifier models to help categorize content, such as news or sports, allowing users to make informed decisions about the data blends they use for training. Jean-Marc Mommessin: Perfect. Is there anything else you’d like to highlight regarding this pipeline? Joey Conway: Yes, I’d like to touch on the Reinforcement Learning (RL) aspect. Following the supervised fine-tuning stage, where we enhanced core skills, we’ve just begun to explore the potential of RL with Nemotron. We believe this will be a significant area of future development. What’s exciting about RL is that its effectiveness is largely tied to the available compute time. The more time we invest, the better the model becomes at specific tasks. In our RL stages, we’ve developed methods to automate the process of asking the model a question, grading its answer, and providing feedback to allow it to learn and improve. You can see on the slide the domains where we’ve applied this: scientific reasoning, instruction following, and chat. If you look at the leaderboards, you’ll see that even with new models emerging, we’ve maintained a strong position in these areas, largely due to the effectiveness of RL in achieving top-tier accuracy. We’re optimistic that we’ll see more of this in the community, with more discussion and publication of techniques and data. We’ve started sharing some of our work in this area and will have much more to come in the next three to six months. Jean-Marc Mommessin: You mentioned RL and instruction following, which ties back to the beginning of our conversation. It seems like you’ve come full circle here. Joey Conway: Exactly. The exciting aspect here is automating the feedback loop wherever possible. For chat, we published a fine-tuned reward model last fall. Those who followed our work might recall that our Llama Nemotron model topped the chat leaderboards then. This was because the reward model provides an automated way to teach the original model whether its responses are good or bad. It essentially grades responses based on helpfulness, conciseness, verbosity, groundedness, and similar factors. This granular feedback per generated response allows the model to improve significantly, often more so than through supervised fine-tuning alone, which typically involves a few passes without a continuous feedback loop. Similarly, for instruction following, we use a verifier and a dataset to teach the model whether it followed instructions well or needs to try again. We’re eager to expand this approach to more domains. We’ve already published datasets related to coding and math since the release of this model a few weeks ago, and these have become popular on Hugging Face. I anticipate significant growth in this area within the community. Jean-Marc Mommessin: Alright, so one of the big innovations here, and you touched upon it, but I want to emphasize it, is the ability to toggle reasoning on and off via the system prompt. This is quite unique, and I’m sure many will follow suit. Could you expand on the idea behind this, how you see it applying to agents and beyond, its value, and the key challenges in implementing it? Joey Conway: The reasoning on and off capability was a core goal from the outset. We observed that models in the community often excelled in either reasoning or non-reasoning tasks, and we wanted to simplify deployment by having a single model that could handle both. We had to determine the best way to teach the model when to reason and when not to, while also providing enterprises with explicit control, as they often have deeper domain knowledge than we do. The motivation behind this is that reasoning generates significantly more tokens, which can lead to higher latency and cost. While crucial for solving complex problems, it’s not always necessary. We wanted to give enterprises the control to balance accuracy with latency and cost, allowing them to decide when to employ reasoning and when to opt for faster, less computationally intensive responses. Initially, we weren’t sure how to achieve this, as it hadn’t been widely implemented in the community. Our approach in the supervised fine-tuning stage was to explicitly teach the model by presenting the same question with two different answers: one with detailed reasoning and one without. This essentially doubled our dataset for this specific purpose. However, the outcome is a single model where users can simply include “use detailed thinking on” or “use detailed thinking off” in the prompt to control the model’s reasoning process. On the training side, this required more effort to teach the model this distinction. What we have today is essentially a v1, and I expect others will follow this approach. We’re also excited about future developments, such as time or token limits for reasoning and more granular controls. I’m optimistic that we’ll see further breakthroughs in this area within the next six to nine months, as the problem-solving power of reasoning is significant, but it comes with trade-offs that the community will continue to refine. Jean-Marc Mommessin: We all know that the real test comes in production. Production environments are sensitive to latency, cost, and while accuracy and reasoning are vital, excessive reasoning can lead to scalability issues and increased latency. The flexibility you’ve introduced is fantastic, and I can see numerous production use cases that will greatly benefit from the ability to control reasoning on a per-query basis. So, when you were developing this model, you aimed to balance accuracy and efficiency. Could you share some insights into how you made these trade-offs, the timeline for building the model and the team involved, and how you determined the optimal compromise between these two critical factors? Joey Conway: Balancing accuracy and efficiency is always a challenge. Our initial goal was to achieve both, which is a difficult undertaking. We started with the “Super” model, which was the most recent Llama 3.1 70B release from Meta, as our baseline for accuracy. We weren’t sure if we could simultaneously improve accuracy and reduce the model size. We found that through our training techniques and distillation process, we could indeed boost accuracy. We even released an initial checkpoint reflecting this. However, we wanted to go further by incorporating strong reasoning capabilities, aiming for state-of-the-art reasoning scores. This is where the SFT and RL stages came in, which required significant time for synthetic data generation since this type of data didn’t exist. During training, we carefully considered the number of epochs for each skill and continuously measured accuracy. Our goal was to improve performance across all six key areas rather than excelling in just a couple. This balancing act took more time as we experimented to find the right combinations. However, we felt it was crucial to ensure world-class performance in these six enterprise-relevant scenarios, including chat and instruction following. For areas like MMLU, we focused on maintaining performance and preventing regression rather than actively trying to improve scores. So, there were definitely priorities and trade-offs involved. Ultimately, we believe these were the right focus areas for our enterprise customers. Jean-Marc Mommessin: You are releasing this model family as part of the open-source community. We’ve discussed the gaps you aimed to address and the unique reasoning on/off feature for production scalability. Could you share your thoughts on how NVIDIA and your team view the role of these models within the broader open-source and LLM ecosystem, especially given your work building upon the Llama base? Joey Conway: NVIDIA has a long history of contributing models to the open-source community. What excites us about Llama is its strong traction with enterprise customers. While NVIDIA Research publishes extensively across various domains, our goal with Llama Nemotron was to build upon Llama’s momentum in enterprise adoption by focusing narrowly on specific areas. The base Llama models already cover many things exceptionally well, so we saw an opportunity to build on top of that and be very targeted in our enhancements. The recent LlamaCon event and Meta’s announcements sound very promising, and we’re excited about Llama 4 and the ongoing work there. Moving forward, we anticipate continuing to identify specific areas where we can add significant value, while Meta continues to build excellent general-purpose models suitable for enterprise production. From our perspective, reasoning will likely remain a key focus, and we’re also excited about Meta’s advancements in this area. Tool calling, instruction following, and chat are also areas we’ll continue to develop. One area we’re particularly interested in exploring is multilingual capabilities. For large enterprises, supporting multiple languages is crucial. While many models handle individual languages well, we aim to focus on a few key languages and ensure world-class accuracy for reasoning, tool calling, and chat within those. This is likely the next major area of expansion for us, beyond the exciting developments in model architectures like Llama 4’s new MoE architecture, which we’re also keen to explore for potential distillation and optimization for NVIDIA GPUs. So, there’s a lot of exciting work ahead. Jean-Marc Mommessin: When you say multilingual, are you thinking of supporting a broad range, like 50 languages, or a more focused set, perhaps around 5 or 10 initially, given the benchmark challenges you mentioned? Joey Conway: We’ll probably start with a more focused set, perhaps around 5 to 10 languages. The challenge is that the community currently lacks comprehensive benchmarks for tasks like reasoning or tool calling across a wide variety of languages. As we develop these multilingual models, we’re also having to create evaluation data simultaneously, which takes time. If those benchmarks were readily available, the process would be smoother. However, we see this as an exciting challenge. Our initial focus will likely be on a smaller set of languages where we can establish strong performance, given the current limitations in community-wide benchmarks. Jean-Marc Mommessin: Let’s shift gears and talk about another state-of-the-art open-source model you recently released: Parakeet TDT 0.6 B parameters, V2. This model has set a new standard for automatic speech recognition (ASR), transcribing one hour of audio in just one second. That’s 50 times faster than other open-source ASR models, and remarkably, it achieves only a 6% word error rate. This is truly impressive. What else would you like to highlight about this model before we discuss the “how” behind its incredible performance? Joey Conway: It’s worth noting that NVIDIA has been working on ASR models for a long time, even before I joined. We’ve also released many open models in this space over the years. The teams working on this are exceptional, and they consistently strive to balance accuracy with latency and throughput. Parakeet V2 is the latest in this line of high-performance models from NVIDIA. Jean-Marc Mommessin: It sounds like the advancements will keep coming. So, let’s delve into how you achieved this remarkable performance with Parakeet TDT. What kind of architecture did you use? I understand it’s based on a Fast Conformer architecture with specific optimizations like 8x depth-wise separable convolutional downsampling and limited context attention. Could you explain how you arrived at this approach and whether these optimizations primarily enhance speed and throughput or if they also contribute to accuracy and the ability to process long audio segments like a full hour in one shot? Joey Conway: Yes, we’ve explored various architectures for ASR over the years, and the Conformer architecture, originally from Google, has shown great promise. Our goal with Parakeet TDT was to take the Conformer architecture and make it significantly more efficient and faster without sacrificing quality. We’ve implemented several key optimizations.  First, as you mentioned, the depth-wise separable convolution downsampling. At the input stage, we significantly downsample the audio, which reduces the computational cost and memory requirements for processing. Second is the limited context attention. By focusing on smaller, overlapping chunks of audio, we can maintain accuracy while achieving a speedup in processing. Third, on the encoder side, we also utilize a sliding window attention technique, which allows us to process longer audio files without having to split them into shorter segments. This is crucial for handling long-form audio like a full hour in a single pass. Beyond the Conformer architecture, Parakeet TDT incorporates a Token and Duration Transducer (TDT). Traditional Recurrent Neural Network (RNN) transducer technology processes audio frame by frame. What we’ve done with TDT is enable the model to predict both the tokens and the expected duration of those tokens. This allows it to make decisions to skip over redundant frames, significantly speeding up the transcription process. This TDT innovation alone contributes to around a 1.5 to 2x speedup. So, there’s a combination of architectural choices and specific optimizations that contribute to Parakeet TDT’s impressive speed and accuracy. Jean-Marc Mommessin: I want to go back to one or two of those. Those are amazing, frankly. The speed increase is remarkable. Joey Conway: Yes, and we have another technique called a label looping algorithm. Essentially, when we’re doing batch inference, this algorithm allows us to advance the tokens independently for different samples. This separation of the workflow enables us to sweep and loop over frames and labels more efficiently, significantly speeding up the decoding process. Lastly, on the decoder side, we’ve moved some of the computation into CUDA graphs, which is a more efficient way to run many small kernels. This optimization alone provided around a 3x speed boost. So, as you can see with TDT models, we’ve been able to achieve speeds comparable to Connectionist Temporal Classification (CTC) decoders, which are also known for their speed, while maintaining high accuracy. Our overall theme is always to balance speed improvements with maintaining or even enhancing accuracy. Techniques like CTC decoders have been around for a while and are fast but might not be as accurate. It really depends on the use case, but we’re always striving for that balance. Jean-Marc Mommessin: Can we revisit the limited context attention? Do you see this technique having broader applications in other areas down the line? Joey Conway: Yes, I believe so. Patterns like the sliding window attention are already used in other areas, such as LLMs. Our research teams are constantly experimenting, looking at successful techniques from different domains, and trying to apply them in new ways. Interestingly, some of the researchers who worked on Parakeet TDT also work on Llama Nemotron, so there’s a cross-pollination of ideas. I do expect that some of these techniques will find broader applications going forward. We also anticipate further improvements to TDT and the Conformer architecture, as we’ve been working on them for several years now. I don’t see these core technologies going away anytime soon; we’ll likely continue to refine them. Jean-Marc Mommessin: Leaving the TDT aside, do you see other potential applications for the Token and Duration Transducer concept in other domains? Joey Conway: That’s a good question. I’m not immediately seeing a direct application of the TDT concept outside of ASR. Its history is rooted in RNNs and RNN transducers, which have primarily been used in speech recognition. However, some of the underlying techniques we’ve applied to it, like using CUDA graphs for optimizing kernel execution, are general techniques that we use whenever we identify bottlenecks in a model’s pipeline. So, while the TDT itself might be domain-specific, some of the optimization strategies we’ve employed could certainly translate to other areas, including large language models. Jean-Marc Mommessin: let’s talk about data. AI data is always a key topic. How do you ensure that the data used to train Parakeet TDT is diverse enough to handle various accents, dialects, vocal ranges, pitches, and noisy background conditions, which often negatively impact ASR performance? Joey Conway: You’re absolutely right. As humans, we naturally filter out accents and background noise to understand speech. However, deep learning models are only as good as the data they’re trained on. Early on, limited data for specific accents or languages resulted in poor performance for those variations. What might have initially seemed like edge cases have become increasingly common, highlighting the need for more representative data. We’ve invested significant effort in curating our datasets to reflect this real-world diversity. We use techniques like classifiers to analyze our data and understand the distributions of accents, dialects, and acoustic conditions. We’ve worked with customers like YUM! Brands, who have drive-through use cases with significant highway noise, illustrating the importance of training the model to handle such challenging environments. Ensuring the right blend and distribution of these conditions in our training data is crucial for the model’s robustness. I’m also excited to announce that we plan to open-source a substantial speech dataset, around 100,000 hours, where we’ve meticulously performed this kind of curation. This dataset will include variations in sound levels, signal-to-noise ratios, background noise types, and even telephone audio formats relevant for call centers. Our goal is to provide the community with high-quality, diverse data that enables models to perform well across a wide range of real-world scenarios. Jean-Marc Mommessin: That’s fantastic news about the open-sourcing of the speech dataset! My final question regarding the Parakeet family: you currently have the 600 million and 1.1 billion parameter models. How do you envision future development for this family? What are the potential directions? Joey Conway: We’re considering development along two main dimensions: model size and the number of supported languages. In terms of size, we’ve released models at the smaller and mid-range to demonstrate the potential, similar to our approach with Llama Nemotron Super. We plan to explore larger models, potentially around 2 billion parameters, which we anticipate will handle even more languages and dialects. On the smaller end, we’re even considering models down to around 50 million parameters. The motivation here is to address use cases at the edge where a smaller footprint is necessary, such as enabling real-time audio processing for robots in noisy environments. We’ll be exploring the right trade-offs for such applications. Technologically, we plan to work on streaming capabilities for TDT. Currently, much of the processing is done in an offline batch mode, but we want to enable real-time, live transcription. And as mentioned, we’re excited about releasing the large, curated speech dataset. Finally, for those looking to deploy these models in production, we recommend exploring techniques like word boosting, which allows for customization of text normalization to include domain-specific terms and acronyms. We aim to provide a wide range of options for users to get started and tailor the models to their specific needs. Jean-Marc Mommessin: I’m very familiar with the NVIDIA Orin platform. Would these Parakeet models currently run on NVIDIA Orin? Joey Conway: Yes, I believe the 0.6 billion parameter model likely would run on Orin. I would need to double-check the exact specifications, but I’m quite confident it’s feasible. Jean-Marc Mommessin: Orin packs a significant punch. I especially love the robotics use case you mentioned. While there’s been a lot of focus on robot vision, the ability to hear and understand quickly is equally crucial, especially for safety. A model that’s 50 times faster and highly accurate in understanding another modality seems like a perfect fit for robotics. Joey Conway: Yes, and the slight hesitation I had earlier was due to the understanding that in robotics, there are often multiple models running simultaneously, including vision models. So, resource allocation is a consideration. However, our push towards smaller, more efficient models is precisely to address these kinds of multi-modal edge computing scenarios. The low latency and real-time processing capabilities of Parakeet are indeed very beneficial for enabling robots to react quickly and safely to auditory cues. Jean-Marc Mommessin: Anything else you’d like to add as a final thought on the Llama Nemotron Ultra and Parakeet families? They’re both open-source, fast, high-throughput, cost-efficient, and run on smaller footprints – are these the key takeaways? Joey Conway: Yes, that’s a great summary. Those were the core objectives we set out to achieve. We aimed for state-of-the-art accuracy, optimized footprints for efficient GPU utilization in terms of latency and throughput, and a commitment to open-sourcing everything to empower the community. We’ve strived to be as community-friendly as possible by releasing datasets, using permissive licenses, and making it easy for people to experiment. We’re eager to see the community’s feedback and the innovative applications they build upon our work. We’re also looking forward to learning from their experiences. Jean-Marc Mommessin: Where are all these models and datasets available? Joey Conway: Everything we’ve published is on Hugging Face – the models and the datasets. The software stack to run them comes from NVIDIA and is available on NGC, our content repository. Much of the underlying software is also open-source and can be found on GitHub. We also provide pip wheels for easier installation. The Nemo framework is the central hub for much of this software stack, whether you want to run the models or fine-tune them. We’ve tried to make it as user-friendly as possible. We use the same software internally to build the models, so it should be relatively straightforward for others to pick up and deploy as well. Jean-Marc Mommessin: Well, Joey, this has been fantastic. I’m continually impressed by NVIDIA’s commitment to giving back to the community with state-of-the-art models that will undoubtedly find their way into production. Thank you so much for your time and insights. I look forward to our next conversation. Joey Conway: Thank you, Jean-Marc. It was my pleasure, and we appreciate the opportunity.  Jean-marc MommessinJean-marc is a successful AI business executive .He leads and accelerates growth for AI powered solutions and started a computer vision company in 2006. He is a recognized speaker at AI conferences and has an MBA from Stanford.Jean-marc Mommessinhttps://www.marktechpost.com/author/jean-marc0000677/NVIDIA AI Releases HOVER: A Breakthrough AI for Versatile Humanoid Control in RoboticsJean-marc Mommessinhttps://www.marktechpost.com/author/jean-marc0000677/Speech-to-Speech Foundation Models Pave the Way for Seamless Multilingual InteractionsJean-marc Mommessinhttps://www.marktechpost.com/author/jean-marc0000677/Lowe’s Revolutionizes Retail with AI: From Personalized Shopping to Proactive Customer AssistanceJean-marc Mommessinhttps://www.marktechpost.com/author/jean-marc0000677/Google DeepMind’s Gemini Robotics: Unleashing Embodied AI with Zero-Shot Control and Enhanced Spatial Reasoning

Ubisoft has revealed updated financial details about its planned joint spinoff subsidiary it's planning with Tencent. In today's full-year financial report for the fiscal year 2024-2025, the company revealed that Tencent will be investing €1.16 billion (about $1.8 billion) in the venture to acquire a 25 percent economic interest in the company. Ubisoft will in turn be pocketing "at least" €500 million (about $559 million) of that transaction, with the remainder serving as "sufficient working capital" for the new company.The news comes on the back of a brutal financial year for Ubisoft. The company reported a 20 percent decline in net bookings, only pulling in €1.8 billion during the year. Digital net bookings also dropped 20 percent to about €1.6 billion, and the normally-robust back catalogue bookings slipped 13.5 percent to €1.3 billion.All told, Ubisoft posted a loss on IFRS operating income of €82.6 million. It reassured its investors that it's still seeing "strong activity metrics" across the Console and PC marketplaces—but the Tencent-backed subsidiary is the center of its transformation plan.Some other tantalizing details of the transaction—expected to wrap up by the end of 2025—also help us assess Ubisoft's outlook. As previously reported, the new subsidiary will be granted licenses and proprietary rights to the Tom Clancy's Rainbow Six, Assassin's Creed series. In exchange, Ubisoft will take in a revenue-based royalty while the subsidiary takes on the development risks. Ubisoft estimates the subsidiary's annual revenue will "exceed €80 million" in the coming years, and that the company's combined value will clock in at a minimum of €5 billion.Related:Ubisoft will also offer "specific shared services" to the spinoff company, indicating there will still be operational ties between "Nubisoft" (that's our nickname for it, not theirs), and the company that spawned it.For those just joining us, Ubisoft arrived at this rocky financial position after a number of high-profile fumbled releases across its many franchises. The Assassin's Creed series has stayed strong, with Assassin's Creed Valhalla, Mirage, and Shadows all finding an audience. But Avatar: Frontiers of Pandora, Star Wars Outlaws, and XDefiant have all been heavy losses for the company. Outlaws' underperformance was so rough it forced Ubisoft to finally release games day-and-date on Steam to boost its PC presence.Frustratingly, Ubisoft still won't say how many copies of Assassin's Creed Shadows it's sold. It stated that the game delivered the "second-highest day-one sales revenue in franchise history," and it was Ubisoft's best day-one performance on the PlayStation store. The best metric it's offered is that players have spent a cumulative 160 million hours stabbing their way through feudal Japan, but for all we know that could just be 800,000 players each playing 200 hours (that is obviously not close to a realistic number).Related:As its games underperformed, Ubisoft faced pressure from investors to take strong action—at one point even reportedly considering a buyout. The company spent the last year slashing costs and closing studios, putting developers out of work and ending fiscal year 2024-2025 with an ever-decreasing number of employees.Its final headcount stood at 17,782 workers as of March 2025, 1,230 fewer than the year before. That decrease wasn't all driven by layoffs—much of it was done by reducing hiring and not backfilling existing roles—but especially with the major studio closures, it seems those with the least agency over Ubisoft's business decisions paid the highest price.

How co-development companies are navigating a turbulent industry in 2025 GamesIndustry.biz catches up with some of the top companies working in the co-dev space to see how they are faring given current industry headwinds Image credit: Bethesda Softworks Feature by Alex Calvin Contributor Published on May 15, 2025 In one form or another, co-development companies have nearly always been around in the games industry. For as long as people have been making games, there have been companies and developers who have made a living helping out other studios with their projects. For the most part, these companies had a low profile, sitting in the shadows of the developers they were working for; but with the growing scale of modern AAA projects, co-development firms are more visible than ever. "There's a growing awareness and desire among studios worldwide to leverage distributed development and external services," says Jake DiGennaro, the chief revenue officer at Virtuos, which, along with Bethesda, co-developed the recently released The Elder Scrolls 4: Oblivion Remastered. "Some reputable studios that have shipped their own IP are even embracing a co-op model, working on original IP development while offering co-development services to others. This approach has long been a tradition in Japan and is now becoming mainstream in North America and Europe." Alejandro Garcia-Tunon, the head of co-dev at Side, adds: "Co-development definitely remains a key part of the game development ecosystem – if anything, more so now than ever before. An increased number of development specialities, as well as scopes that are well beyond what the game industry has seen in the past, have increased costs. "Having a full-time team of experts in each field you'd need to cover just wouldn't make sense, so bringing in an expert is a great way for studios and publishers to tap into that talent without blowing up their budgets." Helping hand The work that co-development companies are being brought in to assist with varies wildly. Sometimes it's individual disciplines (art and animation are the most popular, according to research from the External Development Summit), but for others it is entire segments of a project. "What we are finding is that the balance of the co-development work is changing from coding or feature-focused 'team augmentation' to fully rounded content teams that can take on a whole section of games or even whole games, like we did with the delivery of Lego Horizon Adventures," explains Ashley Liu, the managing director of Keywords Studios' Create arm. Garcia-Tunon adds: "It runs the gamut from individual specialities, like AI, combat, physics, networking, and UI, to taking on entire chunks of a game like a full biome or boss. Sometimes we will cover different aspects even within a single project. Image credit: Konami "When we worked on Silent Hill 2 Remake with Bloober Team, we worked on two boss fights in very different ways. For the fight with Eddie's character, we were creative consultants, advising how we could lift the level from the previous version and make it more of a pivotal moment in the game. With Maria, the final boss fight, Bloober Team already had a vision in mind which we were able to help execute from a technical perspective." Despite co-development companies becoming a bigger part of the industry, the sector isn't immune to the broader economics of the games market. For a variety of reasons, publishers have chosen to be more conservative in funding projects, and fewer games in development means less work for those whose role it is to help out. "Demand grew very strongly over the past decade as games became larger and more complex – developers required additional support," explains Liu. "In the last 18 months, the volume of work reversed due to the overall contraction of the industry, but projects are starting to come back as the industry realises there's been an overcorrection. Players expecting more engaging content will drive some renewed demand." The turbulence in the market, however, does present something of an opportunity for co-development companies. In an uncertain world where publishers are being more risk-averse, partners that provide a great deal of flexibility can be useful. "There are opportunities for co-development, though they may not look like what we've seen in the past," Liu says. "Specifically, as our clients restructure their businesses, there are now more opportunities for partners to fill resource gaps. "There are also opportunities with clients that perhaps would like us to take over their older titles, especially as their best teams move on to new projects." "The headwinds that we face collectively as an industry have reached a breaking point where we can't continue to make games the way we did 20 years ago" Jake DiGennaro, Virtuos Garcia-Tunon adds: "The last couple of years have been an anomaly, no question, but the uncertainty has been really challenging. The co-dev space has historically been growing year over year because the entire industry has been growing – more games equals more people needing help to make games. Our opportunities have come from our ability to quickly adapt to fit client demands." Some co-development companies view themselves as somewhat indispensable. The fact that games now require so many people, so many hours of work, and so many disciplines means that it is tough to make them without external help. "The headwinds that we face collectively as an industry have reached a breaking point where we can't continue to make games the way we did 20 years ago and expect not to run into a wall eventually," DiGennaro believes. "With the rising costs of production, the economics of creating a successful game have become increasingly difficult. "We have to acknowledge the fact that there's something about the way we make games that needs a paradigm shift. While factors like discoverability, delivery, and distribution all play important roles, the process of how we build games is a critical part of the problem," DiGennaro continues. "More than seeing this as an opportunity for co-development companies, I think it's an opportunity to reshape the collective mentality and methodologies of our industry towards increased sustainability and reduced volatility." As always, big games companies are very anxious about their IP and will often go to great lengths to protect it. Before its official reveal, screenshots and marketing materials for The Elder Scrolls 4: Oblivion Remastered were leaked from Virtuos' backend on its Wordpress. "We have very strict processes and information controls at Side that have been honed over decades of work and thousands of games," Garcia-Tunon says. "But our business relies on our reputation – successful projects build great long-term relationships and recommendations from previous partners are worth their weight in gold." Image credit: Keywords Keywords' Liu adds: "Ensuring there are no leaks has always been a challenge in game development, especially for projects that take years to complete with hundreds of people working on it at different times. As with any studio, team composition changes throughout the project lifecycle, making leakages an ongoing area to monitor. "We can leverage experienced global IT and InfoSec departments to protect our client's IP and data. They oversee hundreds of projects, thousands of configurations, and have successfully dealt with a vast number of incidents, allowing us to bring a level of protection surpassing many of our competitors," Liu continues. "In the unlikely event of a data breach, we inform the client first and foremost and together we find a way to resolve the situation. Sometimes, because we are open and honest, the clients trust us more." What comes next Looking to the future, it's clear that co-development isn't going anywhere anytime soon. Modern video games are, by and large, too big and require too much specialist expertise to be made by one company, at least within a reasonable timeframe. Co-dev firms are here to stay. "Fundamentally, I think distributed co-development is how games will be – and should be – built moving forward," DiGennaro says. "It has the potential to make a profound impact in the sustainability of this industry." Liu adds: "Despite the layoffs and project cancellations, this industry will grow. We'll work with partners to release more creative content and better experiences, leveraging new technology to create games and mechanics that have never been seen before. We'll reach new artistic heights that truly engage players." But, as with the rest of the industry, change may be afoot in the near future. New technologies do look like they could have an impact on how games are made. "Workflows are already evolving to include developing technologies like AI," Garcia-Tunon admits. "That said, my outlook is positive – as long as we have humans playing games, we'll need humans to develop and discuss what games should look like."

Christine Farro has cut back on the presents she sends her grandchildren on their birthdays, and she’s put off taking two cats and a dog for their shots. All her clothes come from thrift stores and most of her vegetables come from her garden. At 73, she has cut her costs as much as she can to live on a tight budget.But it’s about to get far tighter.As the Trump administration resumes collections on defaulted student loans, a surprising population has been caught in the crosshairs: Hundreds of thousands of older Americans whose decades-old debts now put them at risk of having their Social Security checks garnished.“I worked ridiculous hours. I worked weekends and nights. But I could never pay it off,” says Farro, a retired child welfare worker in Santa Ynez, California.Like millions of debtors with federal student loans, Farro had her payments and interest paused by the government five years ago when the pandemic thrust many into financial hardship. That grace period ended in 2023 and, earlier this month, the Department of Education said it would restart “involuntary collections” by garnishing paychecks, tax refunds and Social Security retirement and disability benefits. Farro previously had her Social Security garnished and expects it to restart.Farro’s loans date back 40 years. She was a single mother when she got a bachelor’s degree in developmental psychology and when she discovered she couldn’t earn enough to pay off her loans, she went back to school and got a master’s degree. Her salary never caught up. Things only got worse.Around 2008, when she consolidated her loans, she was paying $1,000 a month, but years of missed payments and piled-on interest meant she was barely putting a dent in a bill that had ballooned to $250,000. When she sought help to resolve her debt, she says the loan company had just one suggestion.“They said, ‘Move to a cheaper state,'” says Farro, who rents a 400-square-foot casita from a friend. “I realized I was living in a different reality than they were.”Student loan debt among older people has grown at a staggering rate, in part due to rising tuitions that have forced more people to borrow greater sums. People 60 and older hold an estimated $125 billion in student loans, according to the National Consumer Law Center, a six-fold increase from 20 years ago. That has led Social Security beneficiaries who have had their payments garnished to balloon by 3,000% over the same period, according to the Consumer Financial Protection Bureau.An estimated 452,000 people aged 62 and older had student loans in default, according to a January report from CFPB.Debbie McIntyre, a 62-year-old adult education teacher in Georgetown, Kentucky, is among them. She dreams of retiring and writing more historical fiction, and of boarding a plane for the first time since high school. But her husband has been out of work on disability for two decades and they’ve used credit cards to get by on his meager benefits and her paycheck. Their rent will be hiked $300 when their lease renews. McIntyre doesn’t know what to do if her paycheck is garnished.She floats the idea of bankruptcy, but that won’t automatically clear her loans, which are held to a different standard than other debt. She figures if she picks up extra jobs babysitting or tutoring, she could put $50 toward her loans here and there. But she sees no real solution.“I don’t know what more I can do,” says McIntyre, who is too afraid to check what her loan balance is. “I’ll never get out of this hole.”Braxton Brewington of the Debt Collective debtors union says it’s striking how many older people dial into the organization’s calls and attend its protests. Many of them, he says, should have had their debts cancelled but fell victim to a system “riddled with flaws and illegalities and flukes.” Many whose educations have left them in late-life debt have, in fact, paid back the principal on their loans, sometimes several times over, but still owe more due to interest and fees.For those who are subject to garnishment, Brewington says, the results can be devastating.“We hear from people who skip meals. We know people who dilute their medication or cut their pills in half. People take drastic measures like pulling all their savings out or dissolving their 401ks,” he says. “We know folks that have been driven into homelessness.”Collections on defaulted loans may have restarted no matter who was president, though the Biden administration had sought to limit the amount of income that could be garnished. Federal law protects just $750 of Social Security benefits from garnishment, an amount that would put a debtor far below the poverty line.“We’re basically providing people with federal benefits with one hand and taking them away with another,” says Sarah Sattelmeyer of the New America think tank.Linda Hilton, a 76-year-old retired office worker from Apache Junction, Arizona, went through garnishment before COVID and says she will survive it again. But flights to see her children, occasional meals at a restaurant and other pleasures of retired life may disappear.“It’s going to mean restrictions,” says Hilton. “There won’t be any travel. There won’t be any frills.”Some debtors have already received notice about collections. Many more are living in fear. President Donald Trump has signed an executive order calling for the Department of Education’s dismantling and, for those seeking answers about their loans, mass layoffs have complicated getting calls answered.While Education Secretary Linda McMahon says restarting collections is a necessary step for debtors “both for the sake of their own financial health and our nation’s economic outlook,” even some of Trump’s most fervent supporters are questioning a move that will make their lives harder.Randall Countryman, 55, of Bonita, California, says a Biden administration proposal to forgive some student debt didn’t strike him as fair, but he’s not sure Trump’s approach is either. He supported Trump but wishes the government made case-by-case decisions on debtors. Countryman thinks Americans don’t realize how many older people are affected by policies on student loans, often thought to be the turf of the young, and how difficult it can be for them to repay.“What’s a young person’s problem today,” he says, “is an old person’s problem tomorrow.”Countryman started working on a degree while in prison, then continued it at the University of Phoenix when he was released. He started growing nervous as he racked up loan debt and never finished his degree. He’s worked a host of different jobs, but finding work has often been complicated by his criminal record.He lives off his wife’s Social Security check and the kindness of his mother-in-law. He doesn’t know how they’d get by if the government demands repayment.“I kind of wish I never went to school in the first place,” he says. Matt Sedensky can be reached at msedensky@ap.org and https://x.com/sedensky —Matt Sedensky, AP National Writer

After Kingdom Hearts: Missing Link was canceled earlier this week, fans joked that within it was the most information we’ve gotten on the development of Kingdom Hearts 4 since it was revealed in an announcement trailer three years ago. Square Enix said it is “hard at work” on Kingdom Hearts 4, which will continue Sora, Donald, and Goofy’s adventures in a new story arc. The same day that it announced Missing Link’s cancellation, Square Enix casually dropped several new screenshots of Kingdom Hearts 4, causing the fandom to leap for joy. There’s so much to break down in these images that there’s enough to do a deep dive (pun intended) on all of them. Let’s take a look. The above screenshot shows humanoid Heartless that aren’t Neo Heartless — they kinda look like Sora with that hair — but are also more human, maybe because they’re based on Quadratum humans? There are more new Heartless in the back rocking fists the shape of shields. The Form Change ability for the keyblade we saw from the trailer (Are those stars paopu fruit? Are these chains… chains of memories?) is being used to deal with the new group of Heartless instead of using it to travel as previously shown. According to Tetsuya Nomura, a new Build mechanic in the command menu (which is themed after Quadratum) will allow Sora to scrap and rebuild things. There’s also a Japanese-language version of this exact moment posted on the Japanese Kingdom Hearts social media account that shows it from another angle. In this image, we can see a third creature with long legs, a bell-shaped waist, and a long neck with a spike protruding from it. It looks more like the Pokémon Magearna and Alolan Exeggutor than a Heartless. TheGamersJoint (one of the best Kingdom Hearts YouTubers) noticed that the health and MP UI have changed to move the EX Gauge from on top of the MP bar to curve around the health bar, similar to the original Kingdom Hearts interface. He also notes that the EX gauge will most likely be renamed, as Square Enix usually does, often using EX as a placeholder. Fun fact: Sora has a smartwatch on! What does it do? No one knows. Strelitzia from Kingdom Hearts Cross Union is shown, which showcases an improved character model compared to she looked in the Kingdom Hearts 4 reveal trailer. Also, phones make a return to the series since Kingdom Hearts 3’s introduction of Gummiphones, but this time it’s a rotary phone. Will whippersnappers like Sora even know what it is?  To our Kingdom Hearts community,We’re currently working hard on Kingdom Hearts IV and will continue pouring ourselves into the game’s development. We’re dedicated to making this an experience that lives up to your expectations! [1/2] pic.twitter.com/4lbg6l2ErZ— KINGDOM HEARTS (@KINGDOMHEARTS) May 15, 2025 We get to play as Mickey in what would appear to be a minigame (the command menu is extremely limited, leading us to believe it’s a small gameplay section). It’s been theorized this library could be within Scala ad Caelum, since Mickey was previously tasked to look into the Foretellers from Cross Union last we saw him.  The next image showcases what would appear to be a Darkside Heartless from the original trailer, albeit with a new design that includes wings and an even larger presence. But there are two forms of light rushing toward it. Could it be our hero joined alongside another? Yozora perhaps? Only time will tell. The next photo shows Sora standing in the street in the dark, with sparkles surrounding him on the ground. Some believe the sparkles to be rain, but it could also be broken glass from the surrounding cars following a battle. In the distance is a bright light, which suggests things haven’t settled down in the streets. The last image looks to be from this same scene, but from the perspective of Sora, as the light in the distance illuminates his face. Although there’s no release date or news on a new trailer, Square Enix promises it’s hard at work developing Kingdom Hearts 4. After so many years of radio silence, and in the wake of canceling a game, this was the best move the team could have made to show the fans good faith that their long wait will indeed pay off.

After teasing some news on its social channels earlier this week, Atlus has finally announced that a new Persona game is coming out. No, it’s not Persona 6, but rather Persona 5: The Phantom X, a spin-off that is finally making its way out of China and into other territories.Suggested Reading6 Things To Know Before Starting Persona 5 Tactica Share SubtitlesOffEnglishview videoSuggested Reading6 Things To Know Before Starting Persona 5 Tactica Share SubtitlesOffEnglishUnlike the rest of the Persona 5 games, The Phantom X doesn’t follow Joker, Ryuji, and the rest of the original game’s cast. Instead, it follows a different group of Phantom Thieves led by a high school student code-named Wonder. The team attends a different high school and is accompanied by a different mascot character, the tonfa-wielding owl named Lufel. The other major difference between The Phantom X and most Persona games is that it has microtransactions and gacha mechanics. Yes, to take part in the game’s social sim and dungeon crawling, you’ll have to either wait for a set amount of time to pass or use currency to replenish your energy. Otherwise, the game is free-to-play. Check out the announcement trailer below:Atlus WestPersona 5: The Phantom X will launch on PC and mobile devices on June 26. When Atlus first teased the announcement, some people thought it was referencing Persona 6. However, the RPGs typically shed their previous color scheme and aesthetics with each mainline entry, so most fans knew the black-and-red graphic indicated that whatever this game was, it was still part of the Persona 5 subseries. This, coupled with the owl emoji used on the social posts, clearly hinted at a Phantom X announcement, but the gameThe Phantom X was originally announced in 2023 and has been in an open beta phase in the territory since last year.With Persona 6 rumors floating around, it’s entirely possible that The Phantom X could be the final game in the Persona 5 subseries. If that’s the case, it’s kind of a bummer to end this era on a game that doesn’t star the original cast, but Atlus’ attempts at extending their story have been a bit hit or miss.

Author Colour prediction games have gained massive popularity across digital platforms in recent years, especially in regions like India. These games are easy to play, visually engaging, and promise quick results. But there's one question that every player or potential participant asks: Are colour prediction games based on luck or skill? This article dives deep into the mechanics, psychology, and patterns of these games to answer that question thoroughly. Understanding Colour Prediction GamesColour prediction games are a category of online games where players predict the outcome of the next color (often red, green, or violet) that will appear on a screen or wheel. If their prediction matches the result, they win and earn rewards. Many platforms and apps host such games, often labeling them as entertainment or lottery-style games.One such platform making waves is the 82 lottery game, which combines colour prediction with a digital lottery model. This has led to increased searches for 82 lottery download, as users are eager to try their luck or test their predictive skills on these platforms. The Appeal of Colour Prediction Games in IndiaIndia has witnessed a surge in casual gaming, and colour prediction games India has become a trending search term. With the rise of smartphone penetration and cheap internet access, millions are participating in these games, either for fun or for potential monetary gain.Games like lottery 82 not only offer a chance to win but also present an exciting challenge. The minimal entry barrier, instant results, and the possibility of winning make these games a top choice among young players and casual gamers. Are They Games of Luck or Skill?This is the core question — are these games purely chance-based, or do they require a level of skill? The answer isn’t entirely black and white. The Argument for LuckMany players argue that colour prediction games are fundamentally based on luck. Since the outcomes are typically generated by Random Number Generators (RNGs), each result is independent of the previous ones. This makes it almost impossible to accurately predict the next outcome using any pattern or strategy.Moreover, games like 82 lottery game or others on any colour prediction website often operate similarly to online lotteries, where luck plays the dominant role. In such cases, previous results offer little or no clue about the future outcome. The Argument for SkillOn the other hand, some seasoned players believe that there is a skill element involved, especially in understanding patterns, probability, and psychological tactics. These players use data analysis, game theory, and risk management strategies to make predictions that are slightly better than random guessing.Moreover, experienced users who frequently engage with platforms like lottery 82 often claim that consistent gameplay, proper bankroll management, and emotional control contribute significantly to success. Elements of Skill That May Influence ResultsWhile RNG systems are designed to be random, some subtle aspects of gameplay and human behavior can introduce a skill factor:Pattern Recognition: Some platforms may not use perfectly random systems, and users sometimes identify patterns.Probability and Statistics: Understanding odds and probability distribution helps make informed decisions.Risk Management: Smart players avoid all-in bets and stick to calculated risks.Experience: Like poker or fantasy sports, experience might provide an edge over new players.So while luck dominates, there is some room for skill—particularly on platforms where data transparency or algorithmic design allows for analytical prediction. Responsible Gaming Is KeyWhether you're playing the top colour prediction game or downloading the latest 82 lottery download, it’s crucial to play responsibly. These games can be entertaining and even rewarding, but they should not be seen as a reliable income source. Set a budget, know your limits, and avoid chasing losses.Several colour prediction websites now also offer responsible gaming features to help players manage their time and money effectively. This is a positive step, especially in markets like India, where regulatory frameworks are still evolving. Final Verdict: Luck or Skill?In conclusion, while colour prediction games may seem like they involve skill due to the decision-making element, the majority of their outcomes are driven by randomness. Luck plays a significant role, especially when RNG systems are involved. However, with proper strategies and experience, a small skill factor might influence outcomes in a player’s favor.For casual players, games like lottery 82 or the top colour prediction game provide thrilling entertainment. But always remember — treat them as games, not as a source of guaranteed earnings. Keywords Summary:82 lottery game: A popular colour prediction platform.82 lottery download: Refers to downloading the mobile version of the 82 lottery game.lottery 82: Another name or variant of the 82 lottery platform.top colour prediction game: Highlights the most popular platforms in this genre.colour prediction games India: Discusses the rise of such games specifically in the Indian market.colour prediction website: Refers to platforms offering these games online. Final ThoughtsThe rise of colour prediction games reflects our age-old fascination with prediction, chance, and challenge. Whether you're in it for fun or the thrill of winning, always remember the balance between luck and skill. Play smart, stay safe, and enjoy the colourful world of online predictions.

After Kingdom Hearts: Missing Link was canceled earlier this week, fans joked that within it was the most information we’ve gotten on the development of Kingdom Hearts 4 since it was revealed in an announcement trailer three years ago. Square Enix said it is “hard at work” on Kingdom Hearts 4, which will continue Sora, Donald, and Goofy’s adventures in a new story arc.The same day that it announced Missing Link’s cancellation, Square Enix casually dropped several new screenshots of Kingdom Hearts 4, causing the fandom to leap for joy. There’s so much to break down in these images that there’s enough to do a deep dive (pun intended) on all of them. Let’s take a look.Image: Square EnixThe above screenshot shows humanoid Heartless that aren’t Neo Heartless — they kinda look like Sora with that hair — but are also more human, maybe because they’re based on Quadratum humans? There are more new Heartless in the back rocking fists the shape of shields. The Form Change ability for the keyblade we saw from the trailer (Are those stars paopu fruit? Are these chains... chains of memories?) is being used to deal with the new group of Heartless instead of using it to travel as previously shown.According to Tetsuya Nomura, a new Build mechanic in the command menu (which is themed after Quadratum) will allow Sora to scrap and rebuild things. There’s also a Japanese-language version of this exact moment posted on the Japanese Kingdom Hearts social media account that shows it from another angle. In this image, we can see a third creature with long legs, a bell-shaped waist, and a long neck with a spike protruding from it. It looks more like the Pokémon Magearna and Alolan Exeggutor than a Heartless.Image: Square EnixTheGamersJoint (one of the best Kingdom Hearts YouTubers) noticed that the health and MP UI have changed to move the EX Gauge from on top of the MP bar to curve around the health bar, similar to the original Kingdom Hearts interface. He also notes that the EX gauge will most likely be renamed, as Square Enix usually does, often using EX as a placeholder. Fun fact: Sora has a smartwatch on! What does it do? No one knows.Strelitzia from Kingdom Hearts Cross Union is shown, which showcases an improved character model compared to she looked in the Kingdom Hearts 4 reveal trailer. Also, phones make a return to the series since Kingdom Hearts 3’s introduction of Gummiphones, but this time it’s a rotary phone. Will whippersnappers like Sora even know what it is? We get to play as Mickey in what would appear to be a minigame (the command menu is extremely limited, leading us to believe it’s a small gameplay section). It’s been theorized this library could be within Scala ad Caelum, since Mickey was previously tasked to look into the Foretellers from Cross Union last we saw him. The next image showcases what would appear to be a Darkside Heartless from the original trailer, albeit with a new design that includes wings and an even larger presence. But there are two forms of light rushing toward it. Could it be our hero joined alongside another? Yozora perhaps? Only time will tell.The next photo shows Sora standing in the street in the dark, with sparkles surrounding him on the ground. Some believe the sparkles to be rain, but it could also be broken glass from the surrounding cars following a battle. In the distance is a bright light, which suggests things haven’t settled down in the streets. The last image looks to be from this same scene, but from the perspective of Sora, as the light in the distance illuminates his face.Although there’s no release date or news on a new trailer, Square Enix promises it’s hard at work developing Kingdom Hearts 4. After so many years of radio silence, and in the wake of canceling a game, this was the best move the team could have made to show the fans good faith that their long wait will indeed pay off.See More:

Photo Credit: Lava Lava Shark 4G (pictured) carries a 50-megapixel primary rear camera Highlights Lava Shark 5G could come in blue and gold colour options The handset may get a Unisoc T765 SoC The Lava Shark 5G is expected to ship with Android 15 Advertisement Lava Shark 5G is set to launch in India soon. The company has officially teased key features of the handset alongside announcing its launch date. The price range of the phone has been revealed as well. Key details about the upcoming smartphone had leaked recently. It is expected to have a similar design language to the existing Lava Shark 4G, which introduced in the country in March. The 4G variant gets a Unisoc T606 SoC, a 5,000mAh battery with 18W wired charging support, and a 50-megapixel primary rear camera.Lava Shark 5G India Launch: All We KnowThe Lava Shark 5G will launch in India on May 23, the company said in a press release. It is claimed to have an AnTuTu score of over 4,00,000. The phone will support LPDDR4X RAM and will be priced in the country under Rs. 10,000.Lava added that the upcoming Shark 5G handset will be equipped with a 13-megapixel AI-backed main rear camera sensor. It will have an IP54 rating for dust and splash resistance. An earlier leak suggested that the Lava Shark 5G will come with a dual rear camera unit. It will likely be offered in blue and gold colour options. The round LED flash unit placed in the camera island has a circular design around it.A Geekbench listing of the Lava Shark 5G suggests that it will likely be powered by a Unisoc T765 SoC paired with 4GB of RAM. It is expected to support 64GB of onboard storage and run on Android 15 out-of-the-box. The Lava Shark 4G variant is priced in India at Rs. 6,999 for the 4GB + 64GB option. The smartphone carries a 6.7-inch HD+ display with a 120Hz, a 50-megapixel main rear camera sensor, and an 8-megapixel selfie shooter. It is powered by a Unisoc T606 SoC and a 5,000mAh battery with 18W wired charging support. The smartphone arrives with an IP54 dust and splash-resistant rating. KEY SPECSNEWSDisplay 6.70-inchFront Camera 8-megapixelRear Camera 50-megapixelRAM 4GBStorage 64GBBattery Capacity 5000mAhOS Android 14Resolution 720x1612 pixels For the latest tech news and reviews, follow Gadgets 360 on X, Facebook, WhatsApp, Threads and Google News. For the latest videos on gadgets and tech, subscribe to our YouTube channel. If you want to know everything about top influencers, follow our in-house Who'sThat360 on Instagram and YouTube. Sucharita Ganguly Sucharita is a writer with Gadgets 360 and is mostly found playing with her cat in her free time. She has previously worked at breaking news desks across organizations. Powered by coffee, The Beatles, Bowie, and her newfound love for BTS, she aims to work towards contributing to a better media environment for women and queer folk. More

Rest as an Experience in a Space Rich with Symbolism: Insights from the Argentinian Pavilion at the 2025 Venice BiennaleSave this picture!© Federico CairoliThe proposal by Argentinian architects Marco Zampieron and Juan Manuel Pachué for the Argentinian Pavilion at the 19th Venice Architecture Biennale 2025 is clear from the outset: upon entering Siestario, visitors are immersed in a space of dim lighting and evocative sound, and immediately encounter—at the center of the room, stretched across its width, and acting as the undeniable protagonist—a large, inflated pink bag that instinctively invites repose. This is a silobag, an object commonly used in the Argentinian countryside to store grain and a potent symbol of the country’s export-driven economy. In this setting, the silobag becomes more than a spatial intervention; it also introduces a temporal dimension: an invitation to pause and reflect amid the relentless rhythm of the Biennale. Save this picture!Under the motto “Intelligens. Natural. Artificial. Collective”, general curator and Italian architect Carlo Ratti invites participants to explore and debate the many meanings of these terms, setting the stage for a critical rethinking of contemporary architecture. Within this framework, both national pavilions and individual projects and installations engage with the tensions and possibilities that emerge between the natural, the artificial, and the collective—each from their own distinct perspective.Save this picture!Marco Zampieron and Juan Manuel Pachué — Rosario-based architects and co-founders of Cooperativa — were selected through a public competition organized by the Argentine Foreign Ministry, in partnership with the Chamber of Urban Developers (CEDU), with their winning project to represent Argentina in this edition of the Biennale. Related Article Bahrain Wins the Golden Lion for Best National Participation at the 2025 Venice Architecture Biennale Save this picture!With the support of CEDU and Base Proyectos, ArchDaily was present at the inauguration of the Argentine Pavilion on May 9, at the extraordinary Arsenale in Venice. During the first days of the event, we had the opportunity to speak with the authors of Siestario about their experience participating in the Biennale, the creative process of the project, and other aspects related to their practice and understanding of architecture.Save this picture!AD (Paula Pintos): What was your motivation to participate in the competition to represent Argentina at the Venice Architecture Biennale 2025?Marco Zampieron & Juan Manuel Pachué (MZ & JP): As Cooperativa, our group is deeply engaged with competitions. We participate regularly, advocate for them as a valuable tool, and truly enjoy the process they entail. This one, in particular, stood out because, while it’s a national competition, it offers a strong international platform: it’s about representing the country on one of the most important global stages for architecture. Our studio operates as a collaborative collective that expands or contracts depending on the nature of each project. At its core, our practice revolves around competitions. We see them as a way to access commissions without requiring an extensive portfolio, a strong reputation, or an established body of work. That’s why we especially value competitions that are anonymous and transparent — they open the door to meaningful opportunities, such as building a public-scale pavilion with high visibility, through a more horizontal and equitable framework.Save this picture!AD: How does Siestario relate to the theme proposed by the Venice Biennale 2025 — Intelligens. Natural. Artificial. Collective — and what possible interpretations did you envision based on that premise?MZ & JP: These mottos or curatorial texts often tend to be quite ambiguous — and we see that as something positive, because it opens the door to multiple interpretations. With so many people participating in the Biennale through diverse projects, it’s valuable for the thematic framework to be broad: it allows for a wide range of perspectives and approaches. From there, the question of how to represent Argentina naturally arises. Where do you anchor that representation? In our proposal, what comes through is our way of thinking about architecture, our understanding of the Biennale, and our perspective on the collective.Not only were the notions of intelligence, nature, and collectivity present, but there was also something that moved away from the idea of one place, one solution—a concept that truly captivated us. We find the idea of artificial intelligence somewhat challenging... perhaps we’re a bit romantic in that sense; there’s a certain resistance on our part. When we look at some international pavilions, we see a strong emphasis on technology and technical prowess. And while the artificial and the natural are in constant dialogue, what interests us is thinking about the artificial through the lens of local production—through cultural construction. Not as a race toward technological advancement, but as a way of drawing from the depth of knowledge that already exists within our own culture.It is also important that these competitions are resolved quickly, and often one starts from previous ideas or works. In our case, the collective already had a background, a sensitive research on the territory linked to art and certain key elements. This allowed us to connect with the Biennale's themes more deeply in a short time. We see this in other pavilions as well: many have a foundation of prior work and research. In our architecture — and also in art and other disciplines with which we engage — each person brings a background, certain interests, preferences, and tastes. From the place one comes from, what one observes, all of that forms a layer of information from which we project. We try to reflect on this to make it as conscious as possible, so that we can create something that feels close and authentic to us.Save this picture!Representing Argentina also forces us to ask ourselves what we want to say, how we want to speak about who we are. We are interested in a certain metaphor, a certain poetics. We think about the conditions the country is currently going through, both economically and politically, and this image of Argentina resting upon the Argentine countryside emerges, along with the cost of this rest. The proposal of Siestario suggests that the visitor to the Biennale, weary from seeing so much, could climb uphill to a silobag — that iconic symbol of Argentina's agricultural production — and rest, take a nap. But of course, that silobag is the one that stores soybeans, and soybeans, with all their symbolic, productive, and also destructive weight: they contaminate water, devastate forests, and affect local communities. So, Siestario doesn’t just offer a break; it also questions what that rest truly means. There’s a critique, a political stance, a reading on the artificial, on comfort, on the cost of being at ease. All of this emerges when the concepts are linked: the natural, the artificial, the collective. And here, the siesta also appears as a gesture, as a metaphor, as a critique, as a refuge.Save this picture!AD: What motivated you to use the silobag as the central element of the project? How was the process of re-signifying this object within the exhibition context of the Biennale? MZ & JP: The relationship with the silobag is not new to us; in fact, we had already worked with this element in previous projects, documenting and traveling through different towns. This prior research allowed us to see the silobag not just as a functional element, but also as an object loaded with meaning. The process of re-signifying it here at the Biennale involves decontextualizing it. We wanted to change its scale, give it a new role, and use it as a place for a nap, a space for rest. By doing so, the silobag transforms, ceasing to be just an agricultural object and becoming something more playful, connected to the body, fatigue, and the intimacy of the siesta. It’s a simple gesture, but a profound one: it proposes a space for people to recline and feel comfortable in it.Furthermore, the silobag becomes an object that, although it has strong rural and economic connotations, is reinterpreted within the exhibition context, adding a layer of reflection on the everyday and the collective. The intimacy of the siesta becomes social, as the act of resting in a shared space like this creates a collective moment. In the pavilion, we also worked with the atmosphere: the environment is designed to darken, with sounds and lights that induce a sense of relaxation, contributing to creating the right space for that collective rest.Save this picture!AD: What were you hoping to convey through this installation, and how did you expect visitors to experience or interpret the act of "pausing" within the exhibit's journey?MZ & JP: The proposal for this installation revolves around the idea of "stopping." Stopping to think, to look at what you’re doing, to know yourself, to decide. Ultimately, it’s an invitation to stop, to rest, to sleep. We aim for visitors to find that pause within themselves, to disconnect from the fast pace of their surroundings. Furthermore, this act of "pausing" becomes something desirable, almost like an object of desire: the urge to try it, to climb onto the silobag. We wanted to create a physical and experiential experience, not just a contemplative one. Walking through the pavilion becomes an invitation to interact with the space. It even allows walking on the silobag, changing heights, lying down, and resting.We know that at the Biennale, visitors tend to walk a lot and want to see as much as possible. However, this space demands that you stop, that you dedicate time to the experience. It’s not a place to just pass through quickly; it’s a space that asks you to stay for at least 15 or 20 minutes, to climb onto the silobag, to experience the physical effort of doing so, because it’s not a comfortable chair — it’s an experience that requires an investment of time and attention.The curation and design of the rest of the space, along with the projections, are intended so that the installation partially disappears as one moves through it, creating a visual and spatial effect that forces visitors to interact with the place, to stay alert. The project takes a risk; we know that, in the frenetic pace of the Biennale, some may not pay attention to this proposal. But we believe that, to truly enjoy it, it’s necessary to take that time, lie down, listen to the sounds, and be present.Save this picture!AD: How did you approach the relationship between architecture, space, and time within the Argentinian Pavilion?MZ & JP: Our approach has always been to give prominence to the public and collective. In the projects we develop, we try to make the ground floors open and accessible spaces where people can enter spontaneously, without the design being entirely directed toward exclusive use. We are interested in simplicity and clarity, in being radical in the sense of the purity of the idea. This pavilion, for example, uses a single element to express and resolve multiple layers of meaning. The design process seeks to simplify and refine complex ideas, ensuring that the result is direct and clear.In relation to time, we are interested in how architecture, over time, ages and transforms. In Venice, for example, one can see how the city and architecture are not static, how the passage of time causes materials to change, deteriorate, and feel lived in. This concept is present not only in the built architecture but also in the thinking and creation process. For us, time is not just a factor that affects buildings, but a constant in the design process. We value the time spent on physical models, on hand drawing, on processes that endure and allow us to generate something with more depth.Save this picture!Save this picture!AD: There is something disruptive about Siestario: by not displaying so much information or explicit context, the space becomes entirely an experience. What place did you give to the sensory aspect compared to the more direct information exhibition?MZ & JP: This approach is carried over to all our projects, not just the pavilion. We are interested in making spaces sensory experiences, where people not only see but also feel. We seek the space to invite reflection, but also for the enjoyment of the body, for people to engage directly with the environment. In the pavilion, in particular, we wanted the images to function as thresholds, leading you to other places, to the world of the dreamlike. This project, in particular, emphasizes the idea of pause, of stopping to reflect and rest, which connects with our broader vision of how we understand architecture: not just as something built, but as a lived experience.What we aim to create are silent spaces, ones that impose nothing, but at the same time offer power in their simplicity. This is something noticeable in other projects we've done, where the intervention is almost invisible, yet has a significant impact on people's experience, providing a space for them to claim it and inhabit it in their own way.Save this picture!AD: How was the curatorial process in constructing the space through the projections on the sides of the space?MZ & JP:  The curatorial proposal was built with the aim of integrating architecture, art, and landscape, bringing together the different actors. From the outset, we decided to organize the curation of images into three groups: one linked to architecture, another to art, and another to the very recording of the silobag in the field. This division not only provided the visitor with information about the origin and context of the project but also sought to establish a dialogue between architecture and art, something that is fundamental for us. Through the arts, we wanted to tell the story of our landscape and our territory, and we feel that artists have a special sensitivity to read and construct that landscape in a different way.The curatorial line was also guided by an idea that particularly interested us: that from naps emerge dreams, desires, and aspirations of a country. We wanted to propose a positive perspective within a critical reading, showcasing an Argentina that also produces art and architecture. The selected images engage in dialogue with the territory, and they do not respond to a specific temporal cut but rather bring together works from different moments, generating intersections across diverse times. We were interested in showing how these works relate to one another and how, through them, the variable of time appears in architectural production, in the territory, and in the contexts in which these works are being built.Save this picture!Additionally, there was significant archival work: we incorporated original drawings and unpublished material to add historical value. The selection included examples from across the country, even works that are not always associated with formal architecture, such as recreational or leisure spaces. We aimed to represent not only the major cities but also other realities, histories, and geographies. We also included maps and references to historical processes of urbanization and territorial defense, combining them with contemporary works to build the narrative.Save this picture!Save this picture!AD: What does it mean for you to represent Argentina at an event like the Venice Biennale, and how do you experience the responsibility and opportunity that this entails?MZ & JP: Representing Argentina at the Venice Biennale carries significant weight in the national architectural sphere. Winning the competition gave us a certain level of acceptance within the architecture community in Argentina. It’s difficult to fully reflect on the impact of this, as we are still in the process, but we do notice there is recognition. Our motivation was, above all, a genuine interest in participating in the competition, but upon taking on this responsibility, we realized the importance of representing our country at such a prominent event. For our careers, our studio, and our aspirations, this opportunity is significant. Being part of the Venice Biennale is something that makes us feel good because we know that what we do is part of a system that, although imperfect, allows us to reach places like this. It’s an important point in our trajectory and an achievement that fills us with satisfaction.Save this picture! Image gallerySee allShow less About this authorPaula PintosAuthor••• Cite: Pintos, Paula. "Rest as an Experience in a Space Rich with Symbolism: Insights from the Argentinian Pavilion at the 2025 Venice Biennale" [El Pabellón Argentino en la Bienal de Venecia 2025: La siesta como experiencia en un espacio cargado de simbolismos] 15 May 2025. ArchDaily. Accessed . <https://www.archdaily.com/1030139/rest-as-an-experience-in-a-space-rich-with-symbolism-insights-from-the-argentinian-pavilion-at-the-2025-venice-biennale&gt ISSN 0719-8884Save世界上最受欢迎的建筑网站现已推出你的母语版本!想浏览ArchDaily中国吗?是否 You've started following your first account!Did you know?You'll now receive updates based on what you follow! Personalize your stream and start following your favorite authors, offices and users.Go to my stream