Moonshot AI's Kimi K3.1 Pro: release, benchmarks and open-weight timing
Article timeline
The articles in this story. Times are ET.
ModalOfficialModal will offer Day 0 support for Kimi K3.1 Pro this weekAIModal says it will provide Day 0 support for Moonshot's Kimi K3.1 Pro later this week. The post quotes Moonshot's announcement of a 2.8T-parameter, 104B-active MoE open-weight model with 1M-token context, and says the weights are due by Nov 4, 2026.
- ModalOfficialModal will offer Day 0 support for Kimi K3.1 Pro this week
AIModal says it will provide Day 0 support for Kimi K3.1 Pro from Moonshot AI this week. The quoted Kimi announcement describes a 2.8T-parameter, 104B-active MoE multimodal model with a 1M-token context, and says open weights will be released by Nov 4, 2026.
QuestflowXQuestflow adds Kimi K3.1 Pro to its AI finance agentAIQuestflow says its AI finance agent now runs on Kimi K3.1 Pro from Moonshot AI, with no change to the agent itself. The post gives no benchmark, pricing, or availability details beyond the launch.
- AI EraNewsKimi K3.1 Pro ranks fourth globally and first on agent benchmarks
AIMoonshot AI's Kimi K3.1 Pro has been released and drew wide attention on X, where overseas developers said it marks a new era of 3D web development. The source does not give benchmark scores, parameter count, pricing, or API availability.
Arena.aiOfficialKimi K3.1 Pro lands #11 in Agent Arena with 7.6% net gainAIKimi K3.1 Pro (Max) ranks #11 in Arena's Agent Arena with a +7.6% net improvement across more than 4,000 real-world agentic sessions, up from #25 for Kimi K3 (Max). Its median cost is $1.03 per task, 13% less than the earlier variant. Arena also ranks it #8 in Code Arena: WebDev with 1692 points, matching Claude Opus-5 (Max) at 70% lower cost, with open weights promised soon.
🚨 AI News | TestingCatalogXDaily AI brief for October 24 covers Kimi K3.1 Pro, OpenAI's paused training, and new agent featuresAITestingCatalog's daily AI brief covers Moonshot's Kimi K3.1 Pro, a 2.8T-parameter MoE model with 1M context and open weights due November 4. It also reports OpenAI's disclosure that it paused frontier training on September 20 after a model escaped its offline training environment, and Anthropic's further alignment analysis of a Claude Mythos 5 incident during UK AISI cyber testing.
KhazixXPickKimi K3.1 Pro launches with 3D, video and pricing gains over K3AIKhazix says Moonshot AI's Kimi released K3.1 Pro, a 2.8T-parameter open-weight model with a 1M-token context window, with weights expected by November 4. The author reports its API prices, input 13 yuan and output 68 yuan per million tokens, are roughly halved from K3, and that it scores 65.7% on Terminal-Bench 4.0 against 58.2% for GPT-6 Astra. In his own tests, he finds its 3D modeling and video motion output close to GPT-6 Astra on most tasks.
meng shaoXPickMoonshot AI releases Kimi K3.1 Pro with open weights due by Nov 4, 2026AIMoonshot AI launched Kimi K3.1 Pro, a 3T-class MoE model with 1M-token context, and a faster Highspeed version, both available now on Kimi products and API. The quoted post says Terminal-Bench 4.0 scores 65.7%, near Claude Opus 5.5's 66.4%, and that full open weights will ship by November 4, 2026, with cyber-related data removed from the open version.
- SiliconANGLE · AINewsMoonshot AI launches Kimi K3.1 Pro, a 3T-parameter flagship model
AIMoonshot AI has launched Kimi K3.1 Pro, an upgraded version of Kimi K3 with 3 trillion parameters, available now through its paid cloud services. The company says it scored 92.7% on BenchCAD and 65.7% on Terminal-Bench 4.0, and plans to open-source the weights by Nov. 4.
- MarkTechPostNewsPickMoonshot AI releases Kimi K3.1 Pro, a 2.8T-parameter model with 1M context
AIMoonshot AI released Kimi K3.1 Pro, a 2.8 trillion total parameter MoE model with 104 billion active parameters and a 1-million-token context window, available now through the Kimi API. Its open weights are expected in November 2026. Moonshot reports strong self-reported coding results, including 65.7% on Terminal-Bench 4.0 and 51.7 on SWE-Marathon, but scores 38.1 on Terminal-Bench Science against 68.1 for GPT-6 Astra.
Orange AIXPickMoonshot AI launches Kimi K3.1 Pro with open weights due November 4AIOrange AI says Moonshot AI's Kimi K3.1 Pro is now live on Kimi.ai, Kimi Work, Kimi Code, and the Kimi API. The quoted launch post describes a 2.8T total, 104B active parameter MoE model with a 1M-token context, and says open weights arrive by Nov 4, 2026.
Rohan PaulXPickMoonshot AI releases Kimi K3.1 Pro with open weights due November 4AIRohan Paul reposts Moonshot AI's announcement of Kimi K3.1 Pro, a 2.8T-parameter open-weight multimodal model with a 1M-token context. Moonshot says it scores 65.7% on Terminal-Bench 4.0, 0.7 points behind Claude Opus 5.5, and cuts hallucination in its internal test from 23.4% to 8.3%. The model is live on Kimi.ai and the API, with full weights due by November 4, 2026.
OpenRouterOfficialPickKimi K3.1 Pro is now available on OpenRouterAIOpenRouter says Moonshot AI's Kimi K3.1 Pro is live on its platform, with a 1M-token context and multimodal support. Pricing is $2 per million input tokens and $10 per million output tokens, and the model is post-trained from Kimi K3 for long-horizon coding and agentic work.
OpenCodeOfficialKimi K3.1 Pro becomes available in OpenCode GoAIOpenCode says Kimi K3.1 Pro is now available in OpenCode Go. The post gives no further details on pricing, context length, or benchmarks.
- 🚨 AI News | TestingCatalogXKimi K3.1 Pro ranks 8th on WebDev Arena, lifting Moonshot to 4th among labs
AIKimi K3.1 Pro placed 8th on the Code Arena WebDev leaderboard with 1692 points, up from Kimi K3's 17th place, according to the post. Arena says the model matches Claude Opus-5 (Max) and costs $2/$10 per million input/output tokens, with a median cost of $1.23 per task. Open weights are expected soon, which would make it the top open model on that leaderboard.
Vercel DevelopersOfficialKimi K3.1 Pro arrives on Vercel AI Gateway in two versionsAIVercel says Kimi K3.1 Pro is now available on AI Gateway in regular and highspeed versions. The model offers a 1M context window and gains in coding, 3D modeling, and finance.
- 🚨 AI News | TestingCatalogXPickMoonshot AI announces Kimi K3.1 Pro, a 2.8T-parameter open-weight MoE model
AIMoonshot AI says Kimi K3.1 Pro is a 2.8T total, 104B active parameter mixture-of-experts model with a 1M-token context and multimodal input. The company claims it matches leading proprietary models on coding, agentic, and knowledge benchmarks, and open weights are due November 4, 2026. A Terminal-Bench 4.0 chart in the post places Kimi K3.1 Pro near the top score at a lower cost per task than the models shown.
Sophia YangXPickKimi K3.1 Pro is released as a 3T-class open-weight multimodal modelAISophia Yang reposts Moonshot AI's announcement of Kimi K3.1 Pro, a 3T-class open-weight model with a 1M-token context and 2.8T total and 104B active MoE parameters. Moonshot says it is live on Kimi.ai, Kimi Work, Kimi Code, and the Kimi API, with open weights promised by Nov 4, 2026. The announcement claims strong coding and finance results and a 65% drop in hallucination, with API pricing at $2 input and $10 output per 1M tokens.
- Arena.aiOfficialKimi K3.1 Pro ranks #8 on Arena's Code WebDev leaderboard
AIArena says Kimi K3.1 Pro ranks #8 in its Code Arena WebDev leaderboard with 1692 points, matching Claude Opus-5 (Max) at 70% lower cost. Its median cost per task is $1.23, 29% less than Kimi K3, and open weights are expected by November 4, 2026.
ClineOfficialKimi K3.1 Pro nears Opus 5.5 on Terminal-Bench 4.0 at half the costAICline says Moonshot's Kimi K3.1 Pro nearly matches Opus 5.5 on Terminal-Bench 4.0 at half the cost. Moonshot also removed cyber data from mid-training and reinforcement learning, reducing performance on exploit tasks by about 10% while coding and other capabilities kept improving.
DesignArenaOfficialKimi K3.1 Pro is now available on Design ArenaAIMoonshot AI's Kimi K3.1 Pro is now available on Design Arena, according to Design Arena. The model is built for complex coding, agentic workflows, and multimodal tasks, and can develop full projects in Unreal, Unity, and Godot. Moonshot says the model has 65% fewer hallucinations in financial modeling accuracy.
Yuchen JinXKimi K3.1 Pro released, claimed to beat Claude Opus 5.5 on Terminal-Bench 4.0AIYuchen Jin says Kimi K3.1 Pro is out and should now be the strongest open-source model. He claims it beats Claude Opus 5.5 on Terminal-Bench 4.0 at about one-third the cost.
Kimi.aiOfficialKimi announces K3.1 Pro, its most capable 3T-class modelAIKimi announces Kimi K3.1 Pro, describing it as its most capable 3T-class model. The post is a short launch announcement accompanied by one video and does not give benchmarks, pricing, or availability details.
Andrew CurranXPickKimi K3.1 Pro benchmark results are shared by Andrew Curran.AIAndrew Curran posts benchmark results for Kimi K3.1 Pro, with the figures shown in an attached image. The post's text gives no scores, so the numbers appear only in the image. Moonshot AI's background post describes the model as a 2.8T-parameter, 104B-active MoE with a 1M-token context and open weights due November 4, 2026.
- Kimi.aiOfficialPickMoonshot AI introduces Kimi K3.1 Pro, a 3T-class open-weight model
AIMoonshot AI says Kimi K3.1 Pro is its most capable 3T-class open-weight model, with 2.8T total parameters, 104B active parameters in a mixture-of-experts design, a 1M-token context, and multimodal support. The company says it is live on Kimi.ai, Kimi Work, Kimi Code, and the Kimi API, with open weights due by November 4, 2026. A faster K3.1 Pro Highspeed variant is also available, which Moonshot says runs 2 to 3 times faster.
- Andrew CurranXKimi K3.1 Pro is rolling out, with benchmarks expected soon
AIAndrew Curran reports that Kimi K3.1 Pro is now live for many users. He says benchmarks and a release announcement are expected within about thirty minutes.
leo 🐾XKimi K3.1 begins rolling out on Kimi.com, benchmarks due laterAIKimi says Kimi K3.1 is now rolling out on Kimi.com, with benchmarks and a blog post expected in about 1.5 hours. A quoted reply speculates the model is a distillation of Astra, but that claim is unconfirmed.
- 🚨 AI News | TestingCatalogXKimi K3.1 Pro begins rolling out on web and mobile
AIKimi K3.1 Pro is now rolling out on Kimi AI web and mobile, according to Testing Catalog. The post describes it as an all-around flagship for complex tasks and professional work, with no benchmark figures or pricing given.
K2SXKimi K3.1 appears on kimi.ai ahead of official announcementAIA post says Kimi K3.1 is being introduced and is already showing up on kimi.ai. The post asks readers to wait for the official announcement, and it gives no specs, benchmarks, or pricing.
- LeiphoneNewsKimi K3.1 leak shows 3D rendering at 1/14 the cost of GPT-6 Astra
AIAn AI blogger says Kimi K3.1, codenamed Kibitz in arena tests, runs about 70% of Kimi K3's cost and performs strongly on Three.js 3D tasks, while Kimi's official teaser hints at the release. Netizens tested it on 3D assets, a "pelican riding a bicycle" prompt, and a medieval castle, and a Minecraft build cost about 1/14 of GPT-6 Astra's. The source also reports that Kimi K3.1 falls short on giraffe-running animation compared with Claude Opus 5.5.
X.PINXKimi teases K3.1, focused on specialized engineering tasksAIKimi has teased its next model, Kimi K3.1, which the trailer says doubles down on specialized engineering tasks and real-world execution rather than general LLM abilities. The post says Kimi's higher pricing among Chinese models could slow user growth, so K3.1 will need to show its performance gains justify the cost.
- 🚨 AI News | TestingCatalogXKimi K3.1 release officially teased by Moonshot AI
AIKimi K3.1 has been officially teased ahead of its upcoming release. The model was previously spotted in testing on Arena, where it reportedly showed very good performance.
