Corsair Overhauls Prebuilt, 3-Chamber Airflow Case, & Transparent PSUJune 5, 2025Last Updated: 2025-06-05We take a look at Corsair’s upcoming i600 pre-built PC, Air 5400 case, Frame 4000D prototype, and moreThe HighlightsCorsair’s i600 pre-built PC is a new revision on the company’s i500 and overhauls its GPU cooler and CPU radiatorThe Corsair Air 5400 is an airflow-targeted case that has air ducts on the top and bottom of its chassisCorsair has partnered with Singularity to develop the Frame 4000D prototype, which has an interesting power board that handles cable managementTable of ContentsAutoTOC Grab a GN Tear-Down Toolkit to support our AD-FREE reviews and IN-DEPTH testing while also getting a high-quality, highly portable 10-piece toolkit that was custom designed for use with video cards for repasting and water block installation. Includes a portable roll bag, hook hangers for pegboards, a storage compartment, and instructional GPU disassembly cards.IntroWe visited Corsair’s suite at Computex 2025 and liked some of the stuff the company had to show. Editor's note: This was originally published on May 21, 2025 as a video. This content has been adapted to written format for this article and is unchanged from the original publication.CreditsHostSteve BurkeCamera, Video EditingMike GaglioneVitalii MakhnovetsWriting, Web EditingJimmy ThangCorsair Air 5400Corsair will release its Air 5400, which is an airflow-targeted case. On the back side of the case is a giant hole, which couples with a front-mounted radiator that will allow the case to shove air straight out of it. This design allows it to focus air flow for the GPU entirely from its own set of fans at the bottom of the case. This is probably the most interesting case from Corsair we saw at Computex this year. It should be around $220, though that’s dependent on the everchanging tariff situation. Internally, the Air 5400 has a duct at the bottom where the case has 3x120mm fans (the entire case is actually set up to support all 120mm fans, which simplifies things). The duct is there to guide air into the GPU. Corsair claims that the case is getting about a 1-2 degree improvement with the duct in a like-for-like test. If you do end up with a front radiator, then a potential area that gets abandoned in terms of airflow might be around the VRM area and some of the board components like system memory.  There are mounts for fans up on top of the case along with an additional duct. Looking at the back of the case, there are 2 holes on the back, which is surprising for a 120mm fan. The spacing doesn’t look like it would fit a 120mm fan, but Corsair’s plan is to include a bracket that would adapt a 120mm fan here and would actually cut out into the glass area on the back, which would make you lose about 40% of the fan. This should help but raises some questions about whether it may cause acoustic issues when you partially blast air into a glass wall. Speaking of glass, the Air 5400’s glass is laminated. A couple companies are doing this now. Corsair says this helps the glass stay more put together to prevent shattering. Looking at the back side panel, there’s a big acrylic sheet coupled with an area where air can escape. Opening up the back panel, there’s a huge amount of cable-management depth. You can also see that the motherboard tray is punctured all of the way through. This causes concerns around structural rigidity, but Corsair is using a .8mm thick steel, which helps a little bit here. The company has also strengthened the case’s top panel compared to Corsair’s 4000D case, which received negative feedback in that area. The Air 5400 is set to be priced at $220 with 3x120mm fans included. i600 Grab a GN15 Large Anti-Static Modmat to celebrate our 15th Anniversary and for a high-quality PC building work surface. The Modmat features useful PC building diagrams and is anti-static conductive. Purchases directly fund our work! (or consider a direct donation or a Patreon contribution!)Corsair is updating its i500 pre-built PC, which we hated, with its upcoming i600, and the company has improved it a lot. The i500’s GPU cooling solution had basically no contact with any of the power components. The i600, on the other hand, has massive overhauls here. The case itself has only slightly changed, but the changes made accommodate larger radiators. The block for the video card, including the power components, is totally different in a way that looks promising. The stuff that jumped out to us right away about the i600 is the fin stacks for the VRM, which is connected to a shared copper nickel-plated base plate for the GPU and memory. Everything is connected to the same base plate, which is connected to the liquid cooler. This means all of the heat gets dumped into the liquid cooler. There’s pros and cons to this design. The pro is that all of the other components get cooled better. The downside is that the GPU itself is sharing the heat dissipation capacity with all of the other components in the cooler. This means you typically see some increase in the GPU temperature as a result. There’s ups and downs to this approach. It doesn’t necessarily mean one solution is better than the other as long as it’s all cooled. The i600 has copper bars, which contact the MOSFETs. Otherwise, it’s very similar to the i500. Corsair has also modified its CPU cooler radiator, where the company has moved its tanks off to the side. The tubes are also running in a different direction. Corsair is also moving to 25mm thick fans, where previously they had the slimmer 15mm fans. That extra 10mm will help with pressure and performance a lot. We plan on doing a review of the PC as soon as we buy one.Corsair Frame 4000DWe saw a prototype of Corsair’s Frame 4000D, where the company changed a few things. Corsair partnered with Singularity for its powerboard. It’s somewhat similar to Elmore’s BENCHLAB, with the exception of it not logging power. With the case, you basically run all of the power cables into the power board and then route them to their final locations. We count 10 fan headers here along with a bunch of RGB headers. There’s a lot of possibilities with this. Currently, it’s mostly being used as a cable-management tool, but you could, in theory, expand this to include more switches, like fan-control switches. We would really like to see current monitoring. It would make it more expensive but that could be a potentially useful direction to go in where you could monitor on the 12VHPWR, for instance, which would become a great marketing point for Corsair and would be very useful for end users. The front panel is also different as it has a die-cut edge now.  The power supply setup is also different here with Corsair doing an acrylic wall for the PSU instead of steel. The challenge here is that plastic is an incredibly good insulator. This could raise some ESD (electrostatic discharge) concerns and may cause the PSU to lose some of the shielding that steel provides. As a part of this design, Corsair has customized the caps and PCBs so that they get nice color matching. It looks pretty nice. 3D PrintingCorsair was telling us how for its upcoming 4000D and its Frame series cases, it was getting into 3D printable panels and pieces. They showed us how one Corsair employee 3D printed a shroud-like duct, which takes air in through the bottom and shoves it up into where the pump and reservoir are in the image above. These 3D print files are available on Corsair’s account on Printables.  Corsair 5000DWe didn’t care too much about it but Corsair also showed off its new 5000D that has a screen on it, which is a thing companies are doing now. They also had a 5000D case without the screen, which is a larger variant of the Frame 4000D case. Corsair says that it should be priced around $180, but that’s in flux with the tariffs situation.  Corsair Open Concept Visit our Patreon page to contribute a few dollars toward this website's operation (or consider a direct donation or buying something from our GN Store!) Additionally, when you purchase through links to retailers on our site, we may earn a small affiliate commission.Corsair’s open concept at Computex is using some of the same Frame components, where the company is trying to make the Frame series modular and represents an open frame. There’s also an option for fan mounts as well. The company showed a gigantic radiator tower at the show, which is pretty cool to see.

Elon Musk’s sudden breakup with President Trump took over social media on Thursday. But the escalating feud won’t be enough to salvage a Starlink contract with Canada’s Ontario province, which canceled the deal to hit back at Trump’s trade war. Ontario Premier Doug Ford today confirmed to reporters that the Starlink deal is still dead, even though Musk is on the outs with Trump. Ford also couldn’t help but take a shot at the very public demise of the Musk-Trump partnership.“Yes, Starlink is done,” he said. “Are we surprised that they’re not seeing eye-to-eye? I predicted that as soon as that marriage happened. I thought that there would be a divorce real quick.”As for why Ontario won’t revive the deal, Ford indicated that Musk’s past support of Trump and his trade war against Canada went too far. This included Musk tweeting “Canada is not a real country" in response to a petition urging the Canadian government to rescind his citizenship.Recommended by Our Editors“I don’t want to deal with someone who’s attacking our country. And he was one of the number one culprits, Elon Musk. And that’s unacceptable. I can’t do business with someone that’s doing that,” Ford said. SpaceX’s Starlink business was originally supposed to receive a $100 million CAD ($73 million) contract to deploy satellite internet dishes that would connect 15,000 underserved homes and businesses. But in March, Ontario killed the deal to retaliate against Trump’s 25% tariff on Canadian goods. The decision caused anti-Starlink sentiment skyrocketed in the country, but it also undermined Ontario's effort to bring high-speed broadband to rural and remote areas. Still, some Canadian consumers continue to subscribe to and sign up for the satellite internet service, which has been pushing Starlink promotions in the country. In the meantime, Musk’s emerging feud with Trump could trigger some harsh consequences for SpaceX, which receives numerous federal contracts. In a Truth Social post on Thursday, the President said: “The easiest way to save money in our Budget, Billions and Billions of Dollars, is to terminate Elon’s Governmental Subsidies and Contracts.”

Elon Musk is contemplating decommissioning SpaceX's Dragon spacecraft, responding to President Donald Trump's apparent intent to terminate government subsidies and contracts with the billionaire's companies. It looks like the feud between the former allies has quickly turned vicious.SpaceX's CEO initially announced that the company would retire its Dragon spacecraft in an X post on Thursday, with Musk sharing a screenshot of a post published on Trump's Truth Social account earlier in the day."The easiest way to save money in our Budget, Billions and Billions of Dollars, is to terminate Elon’s Governmental Subsidies and Contracts," said Trump in the screenshotted post. "I was always surprised that Biden didn’t do it!" You May Also Like "In light of the President’s statement about cancellation of my government contracts, @SpaceX will begin decommissioning its Dragon spacecraft immediately," Musk wrote on X.SpaceX's Dragon spacecraft are a family of vehicles designed to carry passengers and cargo. The National Aeronautics and Space Administration (NASA) has previously relied upon them to transport astronauts to and from the International Space Station (ISS). Mere hours prior to Musk's announcement, SpaceX posted on X that it was preparing to launch a Dragon next Tuesday.For a few hours, it seemed reasonable to assume that this launch would now not go ahead. However, Musk then appeared to quickly walk back his decision. Responding to an X user advising him to "cool off and take a step back for a couple days," the billionaire subsequently stated that Dragon will not be decommissioned after all.It's unclear whether Musk's initial announcement was sincere, or whether his apparent about-face might be sarcastic. Musk has a history of making flippant comments online with no apparent regard to their consequences. What is clear is that Musk and Trump's relationship is well past the honeymoon phase, and now looks much more like an ugly divorce.If Trump does terminate government contracts with Musk's companies, it would deal a significant blow to the billionaire. According to a Washington Post investigation, NASA has invested over $15 billion in SpaceX alone. When put together with Musk's other companies such as EV automaker Tesla, his various businesses have received at least $38 billion in government contracts, loans, subsidies, and tax credits. Mashable Trend Report: Coming Soon! Decode what’s viral, what’s next, and what it all means. Sign up for Mashable’s weekly Trend Report newsletter. By clicking Sign Me Up, you confirm you are 16+ and agree to our Terms of Use and Privacy Policy. Thanks for signing up! Musk and Trump go through messy public breakup Credit: Roberto Schmidt / AFP via Getty Images Musk's relationship with Trump has significantly deteriorated in recent days. The billionaire announced that he was leaving his position as de facto head of the Department of Government Efficiency (DOGE) last Wednesday, just one day after he criticised Trump's tax bill as undermining its work. The split was presented as amicable at the time, with Trump presenting Musk with a golden key and words of praise. However, their love affair has quickly turned sour.Musk continued to lambast Trump's bill after his departure from DOGE, arguing that it will increase government debt by trillions of dollars. Strongly disagreeing with the president's characterisation of the proposed legislation as a "Big Beautiful Bill," Musk labelled it a "disgusting abomination" and has been calling for lawmakers to crush it.For his part, Trump has claimed that Musk is simply throwing a tantrum because the bill supposedly cut an alleged "EV mandate." The president stated on Thursday that he had asked the billionaire to leave his administration, and that Musk had been "wearing thin.""I took away his EV Mandate that forced everyone to buy Electric Cars that nobody else wanted (that he knew for months I was going to do!), and he just went CRAZY!" Trump claimed. Despite Trump's assertions, he did not abolish any EV mandate as there has never been any U.S. law which makes switching to an electric car mandatory. However, Trump has taken several anti-EV measures since his inauguration, including abolishing incentives encouraging EV adoption, pausing $3 billion in funding for a U.S. charging network, and introducing a $250 annual fee for EV users in his recent tax bill. Related Stories Trump's claim about Musk is an interesting contrast to his statements in March, when he praised the billionaire for not complaining about the supposed end of the non-existent EV mandate. The president made the comments while he and Musk co-hosted a Tesla ad on the White House lawn in an effort to boost the company's cratering stock prices.Tesla's struggling share value has now fallen again amidst Musk's feud with Trump, plummeting more than 14 percent on Thursday to wipe out over $150 billion in value."I don’t mind Elon turning against me, but he should have done so months ago," Trump wrote on Thursday.Meanwhile, Musk went all-in attacking Trump on Thursday, claiming that the president is linked to child sex offender Jeffrey Epstein and sharing posts calling for him to be impeached. Musk has also hit out at Trump's tariffs on international trade, predicting that they will "cause a recession in the second half of the year.""Without me, Trump would have lost the election," Musk alleged on X. "Such ingratitude."

Reddit a porté plainte contre Anthropic devant un tribunal de San Francisco, mercredi 4 juin. DADO RUVIC / REUTERS La plateforme Reddit a porté plainte contre Anthropic, start-up californienne d’intelligence artificielle (IA). Reddit accuse l’entreprise d’avoir utilisé sans autorisation des conversations publiques de ses utilisateurs pour entraîner ses modèles d’IA générative – parmi lesquels figure Claude, un modèle de langage et chatbot concurrent de ChatGPT. D’après la plainte, déposée mercredi 4 juin à San Francisco, Reddit reproche à Anthropic d’avoir entraîné ses modèles de langage à partir de messages « humains » postés sur Reddit. L’action en justice se base notamment sur un document de recherche publié en décembre 2021 par les équipes d’Anthropic et cosignés par le chef de l’entreprise, Dario Amodei. Le texte mentionne certaines conversations spécifiques de Reddit pouvant servir à l’entraînement de tels modèles. Celles tenues sur Wikipédia sont aussi mentionnées. La plainte allègue que malgré les déclarations publiques d’Anthropic, qui a assuré avoir bloqué l’accès à Reddit à ses systèmes automatisés de récolte des données, les robots de l’entreprise ont malgré tout continué « à se connecter aux serveurs de Reddit, et ce plus de cent mille fois » depuis juillet 2024, indique le texte. Lire aussi | Le gouvernement retire une vidéo générée par IA sur la Résistance, à la suite d’une erreur historique Anthropic conteste « Nous ne sommes pas d’accord avec les affirmations de Reddit et nous nous défendrons vigoureusement », a réagi un porte-parole d’Anthropic auprès de l’Agence France-Presse. Fondée à San Francisco par d’anciens ingénieurs d’OpenAI, Anthropic promeut de manière ostensible un développement responsable de l’IA générative. « Cette affaire porte sur les deux personnalités d’Anthropic : la personnalité publique qui tente de faire croire aux consommateurs qu’elle est une entreprise juste qui respecte les limites et la loi, et la personnalité privée qui se moque de toutes les règles qui gênent ses tentatives de s’en mettre plein les poches », accuse Reddit dans la plainte. Reddit demande désormais des dommages et intérêts, et une injonction pour obliger Anthropic à respecter les termes de son contrat d’utilisation. Ce dernier interdit depuis 2024, date d’entrée en Bourse de Reddit, l’utilisation de données issues des discussions Reddit sans qu’un accord, ou un contrat, ne soit signé avec la plateforme. Reddit, qui déclarait en octobre 2024 compter sur 97,2 millions d’utilisateurs actifs chaque jour, a déjà conclu des accords de licence avec d’autres géants de l’IA générative, dont Google et OpenAI. Ces accords permettent à ces firmes d’utiliser les contenus de Reddit suivant des conditions qui protègent les informations confidentielles des utilisateurs, et offrent une compensation financière à la plateforme. Lire aussi | Article réservé à nos abonnés Un pionnier de l’IA veut construire des systèmes non nuisibles à l’humanité Le Monde avec AFP

Join the event trusted by enterprise leaders for nearly two decades. VB Transform brings together the people building real enterprise AI strategy. Learn more Google has released an updated preview of​​ Gemini 2.5 Pro, its “most intelligent” model, first announced in March and upgraded in May, as a preview, intending to release the same model to general availability in a couple of weeks.  Enterprises can test building new applications or replace earlier versions with an updated version of the “I/O edition” of Gemini 2.5 Pro that, according to a blog post by Google, is more creative in its responses and outperforms other models in coding and reasoning.  During its annual I/O developer conference in May, Google announced that it updated Gemini 2.5 Pro to be better than its earlier iteration, which it quietly released. Google DeepMind CEO Demis Hassabis said the I/O edition is the company’s best coding model yet.  But this new preview, called Gemini 2.5 Pro Preview 06-05 Thinking, is even better than the I/O edition. The stable version Google plans to release publicly is “ready for enterprise-scale capabilities.” The I/O edition, or gemini-2.5-pro-preview-05-06, was first made available to developers and enterprises in May through Google AI Studio and Vertex AI. Gemini 2.5 Pro Preview 06-05 Thinking can be accessed via the same platforms.  Performance metrics This new version of Gemini 2.5 Pro performs even better than the first release.  Google said the new version of Gemini 2.5 Pro improved by 24 points in LMArena and by 35 points in WebDevArena, where it currently tops the leaderboard. The company’s benchmark tests showed that the model outscored competitors like OpenAI’s o3, o3-mini, and o4-mini, Anthropic’s Claude 4 Opus, Grok 3 Beta from xAI and DeepSeek R1.  “We’ve also addressed feedback from our previous 2.5 Pro releases, improving its style and structure — it can be more creative with better-formatted responses,” Google said in the blog post.  What enterprises can expect Google’s continuous improvement of Gemini 2.5 Pro might be confusing for many, but Google previously framed these as a response to community feedback. Pricing for the new version is $1.25 per million tokens without caching for inputs and $10 for the output price.  When the very first version of Gemini 2.5 Pro launched in March, VentureBeat’s Matt Marshall called it “the smartest model you’re not using.” Since then, Google has integrated the model into many of its new applications and services, including “Deep Think,” where Gemini considers multiple hypotheses before responding.  The release of Gemini 2.5 Pro, and its two upgraded versions, revived Google’s place in the large language model space after competitors like DeepSeek and OpenAI diverted the industry’s attention to their reasoning models.  In just a few hours of announcing the updated Gemini 2.5 Pro, developers have already begun playing around with it. While many found the update to live up to Google’s promise of being faster, the jury is still out if this latest Gemini 2.5 Pro does actually perform better.  Daily insights on business use cases with VB Daily If you want to impress your boss, VB Daily has you covered. We give you the inside scoop on what companies are doing with generative AI, from regulatory shifts to practical deployments, so you can share insights for maximum ROI. Read our Privacy Policy Thanks for subscribing. Check out more VB newsletters here. An error occured.

Anysphere, the maker of AI coding assistant Cursor, has raised $900 million at a $9.9 billion valuation, Bloomberg reported. The round was led by returning investor Thrive Capital, with participation from Andreessen Horowitz, Accel, and DST Global. The massive round is Anysphere’s third fundraise in less than a year. The 3-year-old startup secured its previous capital haul of $100 million at a pre-money valuation of $2.5 billion late last year, as TechCrunch was first to report.  AI coding assistants, often referred to as “vibe coders,” have emerged as one of AI’s most popular applications, with Cursor leading the category. Anysphere’s annualized revenue (ARR) has been doubling approximately every two months, a person familiar with the company told TechCrunch. The company has surpassed $500 million in ARR, sources told Bloomberg, a 60% increase from the $300 million we reported in mid-April. Cursor offers developers tiered pricing. After a two-week free trial, the company converts users into paying customers, who can opt for either a $20 Pro offering or a $40 monthly business subscription. Until recently, the majority of the company’s revenue came from individual user subscriptions, Bloomberg reported. However, Anysphere is now offering enterprise licenses, allowing companies to purchase the application for their teams at a higher price point. Earlier this year, the company was approached by OpenAI and other potential buyers, but Anysphere turned down those offers. The ChatGPT maker bought Windsurf, another fast-growing AI assistant, reportedly for $3 billion.

Associate Ad Product ProducerGameloftHo Chi Minh City Ho Chi Minh City vn17 minutes agoApplyJob DescriptionAssist in the production and delivery of rich media ads, ensuring projects are completed on time and within scope.Track project progress using tools (e.g., JIRA) and escalate issues to senior producers when necessary.Communicate with external partners to clarify briefs, provide feedback, and ensure timely, high-quality deliverables.Maintain clear documentation and regularly update internal tracking sheets for outsourced work.Review deliverables to ensure they meet creative standards and technical specifications.Collaborate with QA, developers, and cross-functional teams to resolve issues and maintain alignment throughout the project lifecycle.Qualifications1–2 years of experience in digital production, advertising, or game-related project management.Strong organizational skills with attention to detail and the ability to manage multiple projects and deadlines.Proficient in English (written and spoken) and skilled in Google Workspace as well as task tracking platforms (e.g., JIRA,Confluence)Effective communicator and team player, with a proactive mindset and strong problem-solving skills.Familiar with mobile ad platforms and basic technical knowledge of mobile ecosystems; data analysis experience is a plus.Quick to adapt and eager to learn new tools, technologies, and workflows.Additional InformationRecruitment Process:(1) Screening call ( Only candidates whose qualifications align with the role will be contacted. We sincerely appreciate your understanding)(2) Test(3) Interview(s)(4) OfferWork Location and Hours:Work location: 26 Ung Van Khiem, Binh Thanh District, Ho Chi Minh City.Work hours: 08:30 AM - 06:00 PM, Monday - Friday.Why Join Gameloft?You want to work in an exceptional industry and create games downloaded more than 1 billion times per year.You want to be part of a talented and supportive team of pioneers who have a passion for creating video games.You want to contribute to a vibrant and dynamic atmosphere.You want to join a global company and meet great people around the world.You will work on an employment contract with competitive remuneration and benefits package.You are looking for a fun place to work in the heart of the city.What We Offer:A hybrid working model.A dynamic workplace environment, with over 18 nationalities, where hundreds of world-renowned game titles were born.A range of well-being events and policies to support employees physically, mentally, and emotionally.An open-space office, a cafeteria, a terrace and a Gaming Area.Various training packages, including internal training, sponsorship training, and e-learning (iLearn, Udemy,...).Opportunities to collaborate and train with Gameloft worldwide experts and develop yourselves.An attractive monthly salary alongside Tet and other performance bonuses.Minimum 12 days of paid annual leave, plus 5 days of paid sick leave, Christmas leave, Birthday leave,...100% coverage of mandatory insurance and extra healthcare insurance.A monthly allowance to support lunch meals and working-from-home electricity and Internet bills.Other benefits from one of the best employers in Vietnam.What's more?Want to know more? Visit our websites:Company DescriptionJoin The Game!Leader in the development and publishing of multiplatform games, Gameloft® has established itself as a pioneer in the industry, creating innovative gaming experiences for 25 years. Gameloft creates games for all digital platforms, from mobile to cross-platform titles for PC and consoles. Gameloft operates its own established franchises such as Asphalt®, Dragon Mania Legends, March of Empires, and Dungeon Hunter, and also partners with major rights holders including LEGO®, Universal, Illumination, Hasbro®, Fox Digital Entertainment, Lamborghini®, and Ferrari®. Gameloft distributes its games in over 100 countries and employs 3,000 people worldwide. Every month, 1 billion unique users can be reached by advertisers in Gameloft games with Gameloft for brands, a leading B2B offering dedicated to brands and agencies. Gameloft is a Vivendi company. Create Your Profile — Game companies can contact you with their relevant job openings. Apply

Game Science’s long-awaited Xbox Series X/S version of Black Myth: Wukong finally has a release date. It’s launching on August 20th, with pre-orders starting on June 18th at midnight (Beijing Time) at a 20 percent discount. Those who have yet to experience the action RPG on PS5 or PC can avail of its first-ever sale. It runs from June 18th to July 3rd on the PlayStation Store and June 20th to July 11th on Steam and the Epic Games Store, offering the same 20 percent discount. Black Myth: Wukong launched last August and received extensive critical acclaim, including a nomination at The Game Awards 2024 for Game of the Year (it would win Best Action Game and Player Choice). It sold a whopping 18 million units in two weeks and reportedly crossed 25 million as of last February. An expansion is in the works. The Xbox Series X/S version faced its share of delays, with rumors varying between technical issues and an exclusivity deal with PlayStation. While Sony never confirmed this, it did provide “tons of support” to the developer. Game Science’s studio founder also criticized the Series S’s RAM limitations.

One of the key use cases for generative AI involves answering questions over private datasets, with retrieval-augmented generation (RAG) as the go-to framework. As new RAG techniques emerge, there’s a growing need to benchmark their performance across diverse datasets and metrics.  To meet this need, we’re introducing BenchmarkQED, a new suite of tools that automates RAG benchmarking at scale, available on GitHub (opens in new tab). It includes components for query generation, evaluation, and dataset preparation, each designed to support rigorous, reproducible testing.   BenchmarkQED complements the RAG methods in our open-source GraphRAG library, enabling users to run a GraphRAG-style evaluation across models, metrics, and datasets. GraphRAG uses a large language model (LLM) to generate and summarize entity-based knowledge graphs, producing more comprehensive and diverse answers than standard RAG for large-scale tasks.  In this post, we walk through the core components of BenchmarkQED that contribute to the overall benchmarking process. We also share some of the latest benchmark results comparing our LazyGraphRAG system to competing methods, including a vector-based RAG with a 1M-token context window, where the leading LazyGraphRAG configuration showed significant win rates across all combinations of quality metrics and query classes. In the paper, we distinguish between local queries, where answers are found in a small number of text regions, and sometimes even a single region, and global queries, which require reasoning over large portions of or even the entire dataset.  Conventional vector-based RAG excels at local queries because the regions containing the answer to the query resemble the query itself and can be retrieved as the nearest neighbor in the vector space of text embeddings. However, it struggles with global questions, such as, “What are the main themes of the dataset?” which require understanding dataset qualities not explicitly stated in the text.   AutoQ: Automated query synthesis This limitation motivated the development of GraphRAG a system designed to answer global queries. GraphRAG’s evaluation requirements subsequently led to the creation of AutoQ, a method for synthesizing these global queries for any dataset. AutoQ extends this approach by generating synthetic queries across the spectrum of queries, from local to global. It defines four distinct classes based on the source and scope of the query (Figure 1, top) forming a logical progression along the spectrum (Figure 1, bottom). Figure 1. Construction of a 2×2 design space for synthetic query generation with AutoQ, showing how the four resulting query classes map onto the local-global query spectrum.  AutoQ can be configured to generate any number and distribution of synthetic queries along these classes, enabling consistent benchmarking across datasets without requiring user customization. Figure 2 shows the synthesis process and sample queries from each class, using an AP News dataset. Figure 2. Synthesis process and example query for each of the four AutoQ query classes.  About Microsoft Research Advancing science and technology to benefit humanity View our story Opens in a new tab AutoE: Automated evaluation framework  Our evaluation of GraphRAG focused on analyzing key qualities of answers to global questions. The following qualities were used for the current evaluation: Comprehensiveness: Does the answer address all relevant aspects of the question?  Diversity: Does it present varied perspectives or insights?  Empowerment: Does it help the reader understand and make informed judgments?  Relevance: Does it address what the question is specifically asking?   The AutoE component scales evaluation of these qualities using the LLM-as-a-Judge method. It presents pairs of answers to an LLM, along with the query and target metric, in counterbalanced order. The model determines whether the first answer wins, loses, or ties with the second. Over a set of queries, whether from AutoQ or elsewhere, this produces win rates between competing methods. When ground truth is available, AutoE can also score answers on correctness, completeness, and related metrics. An illustrative evaluation is shown in Figure 3. Using a dataset of 1,397 AP News articles on health and healthcare, AutoQ generated 50 queries per class (200 total). AutoE then compared LazyGraphRAG to a competing RAG method, running six trials per query across four metrics, using GPT-4.1 as a judge. These trial-level results were aggregated using metric-based win rates, where each trial is scored 1 for a win, 0.5 for a tie, and 0 for a loss, and then averaged to calculate the overall win rate for each RAG method. Figure 3. Win rates of four LazyGraphRAG (LGR) configurations across methods, broken down by the AutoQ query class and averaged across AutoE’s four metrics: comprehensiveness, diversity, empowerment, and relevance. LazyGraphRAG outperforms comparison conditions where the bar is above 50%. The four LazyGraphRAG conditions (LGR_b200_c200, LGR_b50_c200, LGR_b50_c600, LGR_b200_c200_mini) differ by query budget (b50, b200) and chunk size (c200, c600). All used GPT-4o mini for relevance tests and GPT-4o for query expansion (to five subqueries) and answer generation, except for LGR_b200_c200_mini, which used GPT-4o mini throughout. Comparison systems were GraphRAG (Local, Global, and Drift Search), Vector RAG with 8k- and 120k-token windows, and three published methods: LightRAG (opens in new tab), RAPTOR (opens in new tab), and TREX (opens in new tab). All methods were limited to the same 8k tokens for answer generation. GraphRAG Global Search used level 2 of the community hierarchy. LazyGraphRAG outperformed every comparison condition using the same generative model (GPT-4o), winning all 96 comparisons, with all but one reaching statistical significance. The best overall performance came from the larger budget, smaller chunk size configuration (LGR_b200_c200). For DataLocal queries, the smaller budget (LGR_b50_c200) performed slightly better, likely because fewer chunks were relevant. For ActivityLocal queries, the larger chunk size (LGR_b50_c600) had a slight edge, likely because longer chunks provide a more coherent context. Competing methods performed relatively better on the query classes for which they were designed: GraphRAG Global for global queries, Vector RAG for local queries, and GraphRAG Drift Search, which combines both strategies, posed the strongest challenge overall. Increasing Vector RAG’s context window from 8k to 120k tokens did not improve its performance compared to LazyGraphRAG. This raised the question of how LazyGraphRAG would perform against Vector RAG with 1-million token context window containing most of the dataset. Figure 4 shows the follow-up experiment comparing LazyGraphRAG to Vector RAG using GPT-4.1 that enabled this comparison. Even against the 1M-token window, LazyGraphRAG achieved higher win rates across all comparisons, failing to reach significance only for the relevance of answers to DataLocal queries. These queries tend to benefit most from Vector RAG’s ranking of directly relevant chunks, making it hard for LazyGraphRAG to generate answers that have greater relevance to the query, even though these answers may be dramatically more comprehensive, diverse, and empowering overall. Figure 4. Win rates of LazyGraphRAG (LGR) over Vector RAG across different context window sizes, broken down by the four AutoQ query classes and four AutoE metrics: comprehensiveness, diversity, empowerment, and relevance. Bars above 50% indicate that LazyGraphRAG outperformed the comparison condition.  AutoD: Automated data sampling and summarization Text datasets have an underlying topical structure, but the depth, breadth, and connectivity of that structure can vary widely. This variability makes it difficult to evaluate RAG systems consistently, as results may reflect the idiosyncrasies of the dataset rather than the system’s general capabilities. The AutoD component addresses this by sampling datasets to meet a target specification, defined by the number of topic clusters (breadth) and the number of samples per cluster (depth). This creates consistency across datasets, enabling more meaningful comparisons, as structurally aligned datasets lead to comparable AutoQ queries, which in turn support consistent AutoE evaluations. AutoD also includes tools for summarizing input or output datasets in a way that reflects their topical coverage. These summaries play an important role in the AutoQ query synthesis process, but they can also be used more broadly, such as in prompts where context space is limited. Since the release of the GraphRAG paper, we’ve received many requests to share the dataset of the Behind the Tech (opens in new tab) podcast transcripts we used in our evaluation. An updated version of this dataset is now available in the BenchmarkQED repository (opens in new tab), alongside the AP News dataset containing 1,397 health-related articles, licensed for open release.   We hope these datasets, together with the BenchmarkQED tools (opens in new tab), help accelerate benchmark-driven development of RAG systems and AI question-answering. We invite the community to try them on GitHub (opens in new tab).  Opens in a new tab

This story appeared in The Logoff, a daily newsletter that helps you stay informed about the Trump administration without letting political news take over your life. Subscribe here.Welcome to The Logoff: President Donald Trump and Elon Musk’s increasingly nasty public split is dominating headlines today, and you can read all about it here. But it also means that another change Trump made this week is flying under the radar — one with consequences for millions of people around the world.What just happened? On Wednesday evening, the Trump administration announced a total ban on entry to the US for citizens of 12 countries, resurrecting a key first-term policy. The ban also includes partial restrictions on seven other countries and is set to take effect this coming Monday, June 9.Which countries does the ban impact? Immigrants, students, and tourists will all be affected. Citizens of Afghanistan, Myanmar, Chad, Republic of the Congo, Equatorial Guinea, Eritrea, Haiti, Iran, Libya, Somalia, Sudan, and Yemen are subject to a total ban, while all immigrant visas and many classes of non-immigrant visa for citizens of Burundi, Cuba, Laos, Sierra Leone, Togo, Turkmenistan, and Venezuela are covered by the partial restrictions.What’s the context? This new ban isn’t a surprise — a “Muslim ban,” which was eventually upheld by the Supreme Court after multiple iterations, was a flagship policy in the first Trump administration, and the second Trump administration has previously signaled its intent to bring back some version of the policy. In March, it was reported that the administration was considering various restrictions on at least 43 different countries. What’s different this time? The new ban covers more countries and cites a wider range of justifications. In addition to national security concerns, the ban also alleges “significant risks” of visa overstays from some of the impacted countries. It comes as the administration steps up an ever-more-aggressive campaign of deportation and anti-immigration policies. RelatedThe real reasons Musk is feuding with TrumpAnd with that, it’s time to log off…The NBA finals start tonight, featuring the Oklahoma City Thunder and the Indiana Pacers! I really enjoyed this preview of the series from Defector, which is a great starting point whether or not you’ve been following along through the playoffs. That being said, the Thunder head into the series as favorites — something I, a former Washington state resident, cannot condone. Go Pacers, and long live the Sonics.See More: