Google's Gemini 4 Carbon: cyber defenders first, Fairwind Program expansion, pricing disputed
Article timeline
The articles in this story. Times are ET.
- The DecoderNewsPickGoogle DeepMind releases Gemini 4 Carbon, topping the Artificial Analysis Intelligence Index
AIGoogle DeepMind has released Gemini 4 Carbon, two weeks after Argon, and it scores 59 on version 5.0 of the Artificial Analysis Intelligence Index, ahead of Argon at 50. Independent evaluations report gains in agentic coding, knowledge work and legal research, while hallucination rises from 4 to 11 percent and runtimes and token counts grow in several tests. Access is limited to selected cybersecurity experts at first, with broader developer and consumer availability planned after further feedback and guardrail adjustments.
Arena.aiOfficialGemini 4 Carbon (High) reaches the Pareto frontier in three Arena leaderboardsAIArena says Gemini 4 Carbon (High) from Google DeepMind lands on the Pareto frontier in Agent Arena, Code Arena: WebDev, and Text Arena. In Agent Arena it ranks second with a +16.2% net improvement at a $1.80 median cost per task, 72% less than first-place Claude Opus 5.5 (Max).
GuizangXGoogle releases Gemini 4 Carbon, claiming benchmark leads over rival modelsAIGoogle announces Gemini 4 Carbon, its next frontier model, which scores 59 on the Artificial Analysis Intelligence Index v5.0 and 88.2% on SEC-bench Pro. Guizang reposts Sundar Pichai's announcement and says Carbon scores above Fable 5.1 and Opus 5.5 on nearly all evaluations, with a large lead on Terminal-Bench 4.0.
- Ars Technica · AINewsGoogle announces Gemini 4 Carbon, a frontier model not yet available to users
AIGoogle has announced Gemini 4 Carbon, an improvement on Gemini 4 Argon released two weeks earlier. Google says it gains 89 points on WebDev Arena, ranks fourth on that leaderboard, and is twice as likely as Argon to produce code free of vulnerabilities. Google is rolling it out in stages, and the article notes that its lead over GPT-6 Astra, Claude Opus 5.5, and Claude Fable 5.1 is only a few points.
Elon MuskXGoogle releases Gemini 4 Carbon, ranking first on the AA Intelligence Index v5.0AIGoogle announces Gemini 4 Carbon, its next frontier model, which scores 59 on the Artificial Analysis Intelligence Index v5.0 and ranks first there. It reports 88.2% on SEC-bench Pro, 54.8% on ExploitGym and 71.0% on CWE-bench, and rollout starts with trusted cyber defenders before wider availability. Musk's post only says "Impressive" and reposts Sundar Pichai's announcement.
🚨 AI News | TestingCatalogXGemini 4 Carbon (High) ranks first on Text Arena with 1528 pointsAIGoogle's Gemini 4 Carbon (High) ranked first on Text Arena with 1528 points, ahead of Gemini 4 Argon (High) at 1520. It ranked second on Agent Arena with a +16.2% net improvement at $1.80 median cost per task, and fourth on Code Arena: WebDev with 1768 points. The post is a relay of an Arena announcement that includes the leaderboard figures and a Pareto frontier chart.
elvisXGoogle introduces Gemini 4 Carbon, a successor to ArgonAIGoogle's Logan Kilpatrick introduced Gemini 4 Carbon, which he says is a big jump over Argon released two weeks earlier. The quoted post says Carbon is state of the art on cyber and long-horizon tasks and ranks first on Artificial Analysis V5. An image lists Terminal-Bench 4.0 cheating falling to 0.0% and unsupported claims cut about 2x versus Argon.
- MarkTechPostNewsPickGoogle DeepMind introduces Gemini 4 Carbon for cyber defense and coding
AIGoogle DeepMind has introduced Gemini 4 Carbon, a frontier model it says is built for long-running agentic coding and cybersecurity defense. Access starts with a vetted group of cyber defenders and will reach developers later with no date given, at $2 per million input tokens and $10 per million output tokens during the introductory period. Its benchmark claims are mostly Google's own or self-run, and independent replication is not yet available.
- 🚨 AI News | TestingCatalogXGoogle announces Gemini 4 Carbon, ranked first on Artificial Analysis V5
AIGoogle announced Gemini 4 Carbon, which the author says ranks first on the Artificial Analysis V5 ranking. Google says it is rolling out first to cyber defenders, with 88.2% on SEC-bench Pro, 54.8% on ExploitGym and 71.0% on CWE-bench. The quoted post from Logan Kilpatrick says it is a large jump over Gemini 4 Argon, released two weeks earlier.
Yuchen JinXGemini 4 Carbon beats Opus 5.5 on Artificial Analysis per-task cost, per Yuchen JinAIYuchen Jin says Artificial Analysis reports that Gemini 4 Carbon is both smarter and cheaper per task than Opus 5.5. He notes that the benchmark ran Gemini 4 Carbon through Antigravity CLI, and he plans to test it in Codex and Omnigent himself.
Varun MohanXGoogle's Gemini 4 Carbon ranks first on Artificial Analysis' new indexAIVarun Mohan says Gemini 4 Carbon is already first on Artificial Analysis' new index. He says internal users have found it useful for hard tasks in Antigravity and that it performs strongly on cyber tasks such as finding and patching vulnerabilities. Access is expanding to the Fairwind Program before wider availability in Antigravity.
Rohan PaulXPickGoogle's Gemini 4 Carbon leads several benchmarks against GPT-6 Astra and Claude Opus 5.5AIRohan Paul relays Google DeepMind's announcement of Gemini 4 Carbon, a new frontier model that Google says is state-of-the-art in cybersecurity and long-horizon work. Carbon scores 88.2% on SEC-bench Pro, 54.8% on ExploitGym, 71.0% on CWE-bench, and 96.2% on GraphWalks, against GPT-6 Astra's 71.8% and Claude Opus 5.5's 66.8% on the long-context test.
Josh WoodwardOfficialPickGoogle releases Gemini 4 Carbon, ranking first on the AA Index v5.0AIGoogle introduces Gemini 4 Carbon as its next frontier model, two weeks after Gemini 4 Argon. Carbon scores 59 on the Artificial Analysis Intelligence Index v5.0 and leads cybersecurity evaluations, with 88.2% on SEC-bench Pro. Access begins with trusted cyber defenders, then expands through the Fairwind Program before wider availability.
- 🚨 AI News | TestingCatalogXGoogle announces Gemini 4 Carbon, topping Artificial Analysis V5 ranking
AIGoogle announces Gemini 4 Carbon, which takes the top spot on the Artificial Analysis V5 ranking and is rolling out first to cyber defenders. Per the post, it scores 88.2% on SEC-bench Pro for vulnerability finding, 54.8% on ExploitGym for exploitation, and 71.0% on CWE-bench for patching. Google's Logan Kilpatrick says it is a major jump over Gemini Argon, released two weeks earlier.
- Arena.aiOfficialPickGemini 4 Carbon ranks #1 in Text Arena and #2 in Agent Arena
AIArena says Google DeepMind's Gemini 4 Carbon (High) debuted at #1 in Text Arena with 1528 points, #2 in Agent Arena, and #4 in Code Arena: WebDev. In Agent Arena it shows a +16.2% net improvement over baseline across 21.9K real-world agentic sessions, at a $1.80 median cost per task, 72% less than Claude Opus 5.5 (Max).
Nathan LambertXGoogle announces Gemini 4 Carbon, its next frontier modelAIGoogle says Gemini 4 Carbon ranks first on the Artificial Analysis Intelligence Index v5.0 with a score of 59, two weeks after Gemini 4 Argon. It reports 88.2% on SEC-bench Pro, 54.8% on ExploitGym and 71.0% on CWE-bench, and access will roll out gradually to trusted cyber defenders and partners. The author, Nathan Lambert, says he has not known how to try Gemini in the agentic era.
Artificial AnalysisOfficialPickGemini 4 Carbon tops Artificial Analysis Intelligence Index v5AIArtificial Analysis says Google's Gemini 4 Carbon scores 59 on its Intelligence Index v5, the top score in its benchmark set. The model ranks second to Claude Opus 5.5 on the Coding Agent Index, scoring 59 against 62, and leads AA-Omniscience v1.1 with a score of 54. Gemini 4 Carbon is rolling out to selected users only, with a 50% launch discount and $4/$20 per million input/output tokens at standard pricing.
👩💻 Paige BaileyOfficialPickGoogle releases Gemini 4 Carbon, its next frontier modelAIGoogle introduces Gemini 4 Carbon, two weeks after Gemini 4 Argon, and says it ranks #1 on the Artificial Analysis Intelligence Index v5.0 at 59. The announcement reports 88.2% on SEC-bench Pro, 54.8% on ExploitGym and 71.0% on CWE-bench, and says access will roll out gradually, starting with trusted cyber defenders.
fofrXGoogle announces Gemini 4 Carbon, its newest frontier modelAIGoogle announces Gemini 4 Carbon, a new model it says reflects faster progress at the frontier. The company says its researchers and models are advancing capabilities while strengthening safeguards in tandem. The post gives no benchmark scores, parameter counts, or pricing.
Yi TayXPickGoogle releases Gemini 4 Carbon, claiming top scores on AA Index v5.0AIYi Tay reposts Google's announcement of Gemini 4 Carbon, Google's next frontier model. Google says it scores 59 on the Artificial Analysis Intelligence Index v5.0, ranking first, and reports 88.2% on SEC-bench Pro, 54.8% on ExploitGym and 71.0% on CWE-bench. Access starts with trusted cyber defenders this week, expanding to partners through the Fairwind Program before wider availability.
Dmitry LyalinXGoogle introduces Gemini 4 Carbon, a new modelAIGoogle announces Gemini 4 Carbon in a blog post linked from the post, but the post gives no benchmark, pricing, or availability details. The author says they are eager for broad availability.
- Yuchen JinXGemini 4 Carbon reportedly tops Opus 5.5 and GPT-6 Astra on benchmarks
AIYuchen Jin says Google's Gemini 4 Carbon beats Claude Opus 5.5 and GPT-6 Astra on most benchmarks and ranks first on the Artificial Analysis Intelligence Index v5. He frames it as a possible new Google comeback and questions whether it could affect Anthropic's IPO.
Philipp SchmidOfficialPickGoogle announces Gemini 4 Carbon, a frontier model with stronger coding and cybersecurityAIGoogle says Gemini 4 Carbon, the next iteration of its frontier model, builds on Argon with stronger coding and cybersecurity. It keeps the same $2 per million input and $10 per million output token pricing and scores 59 on the AA v5 index. The model is in testing with trusted partners, and Google says developer access is coming as soon as possible.
Patrick LoeberXGoogle announces Gemini 4 Carbon, its next frontier modelAIin a post pointing readers to its blog for details. The post itself gives no benchmark scores, pricing, or availability information.
Evan OteroXGoogle introduces Gemini 4 Carbon, ranked #1 on Artificial Analysis V5AIGoogle announces Gemini 4 Carbon, which it says reached state-of-the-art results on several benchmarks two weeks after the Argon release. The model ranks #1 on Artificial Analysis V5, and Google says it improves on coding usability and alignment. Google says the model will be available soon.
koray kavukcuogluOfficialPickGoogle previews Gemini 4 Carbon with long-horizon coding and cyber defense gainsAIGoogle DeepMind's Koray Kavukcuoglu says Gemini 4 Carbon is a preview of Google's next frontier model, arriving two weeks after Gemini 4 Argon. He claims state-of-the-art long-horizon coding and cybersecurity defense performance, and reports a score of 59 on Artificial Analysis Intelligence Index v5.0, ranking first overall. Access starts this week with a subset of cyber defenders through the Fairwind Program, with broader access to follow.
Google AIOfficialPickGoogle announces Gemini 4 Carbon with cybersecurity and long-horizon capabilitiesAIGoogle AI announces Gemini 4 Carbon, which it calls its strongest model, with state-of-the-art cybersecurity and long-horizon capabilities. The post says Carbon improves on most benchmarks and shows strong frontier performance in autonomous coding and cyber defense. Access is first limited to trusted cyber defenders in the Fairwind program, with broader access promised as soon as possible.
Demis HassabisOfficialGemini 4 Carbon is state of the art on key benchmarks, Hassabis saysAIDemis Hassabis says Google's Gemini 4 Carbon is state of the art across key benchmarks and combines frontier capabilities with deep safety and security research. The post gives no benchmark names or scores.
GoogleOfficialPickGoogle introduces Gemini 4 Carbon, its latest frontier modelAIGoogle announces Gemini 4 Carbon, a new frontier model it says improves performance across domains. The company reports that it ranks #1 on v5.0 of the Artificial Analysis Intelligence Index and sets a new state-of-the-art in cybersecurity and long-horizon capabilities. It begins rolling out this week to cyber defenders through the Fairwind Program, with expanded access coming soon.
Google DeepMindOfficialPickGoogle DeepMind announces Gemini 4 Carbon as its next frontier modelAIGoogle DeepMind says Gemini 4 Carbon is its next frontier model. The company states that the model is state-of-the-art across cybersecurity and long-horizon capabilities.
Logan KilpatrickOfficialPickGoogle releases Gemini 4 Carbon, a major step up from ArgonAIGoogle releases Gemini 4 Carbon, which it says is a big jump over Argon, released two weeks earlier. Google says the model is state of the art on cyber and long-horizon tasks, ranks first on the Artificial Analysis V5 index, and is more aligned.
Sundar PichaiOfficialPickGoogle releases Gemini 4 Carbon, its next frontier modelAIGoogle introduces Gemini 4 Carbon, its next frontier model, two weeks after Argon. Carbon ranks #1 on the Artificial Analysis Intelligence Index v5.0 at 59 and scores 88.2% on SEC-bench Pro. Access starts with trusted cyber defenders and expands to partners through the Fairwind Program before wider availability.
- Google DeepMindOfficialPickGoogle DeepMind announces Gemini 4 Carbon with staged cyber defender access
AIGoogle DeepMind announces Gemini 4 Carbon, which ranks #1 on the Artificial Analysis Intelligence Index at 59. The model scores 88.2% on SEC-bench Pro, 54.8% on ExploitGym, and 71.0% on CWE-bench, and it is first available to a subset of cyber defenders. It launches at $2 per million input tokens and $10 per million output tokens, with broader developer and consumer access to follow.
- Artificial Analysis ArticlesOfficialPickGemini 4 Carbon ranks first on Artificial Analysis Intelligence Index v5
AIArtificial Analysis says Google's Gemini 4 Carbon scores 59 on Intelligence Index v5, the highest result, and leads four of nine evaluations. It ranks second on the Coding Agent Index, three points behind Claude Opus 5.5 with Claude Code, and costs $7.76 per Intelligence Index task, about 23% less than Claude Opus 5.5. The model is rolling out to selected users only, and Google is offering a 50% launch discount with no confirmed end date.
- AI EraNewsGoogle's Gemini 4 Carbon appears in early blind tests on LMArena
AIGoogle's Carbon model has begun testing, with partners getting early access and LMArena running blind comparisons ahead of release. The source describes the model as Google's strongest yet and says its launch is near, but gives no benchmark scores, parameter count, pricing, or availability details.
- AI EraNewsGoogle's Gemini 4 Carbon model reportedly surfaces in testing
AIAn overseas outlet reports that Google's next-generation model, Gemini 4 Carbon, has quietly begun testing as a confidential model. The source text provided is only a short teaser, so no benchmark scores, parameter counts, pricing, or availability details can be confirmed from it.