Moonshot Releases Kimi K3.1 Pro, Debuts 4th on Intelligence Index at $2/$10 Pricing
Moonshot released Kimi K3.1 Pro, a multimodal open-weight model post-trained from Kimi K3 for long-horizon coding and agentic work. It debuted 4th on the Artificial Analysis Intelligence Index with a score of 55, an 11-point gain over Kimi K3, trailing Claude Opus 5.5, Claude Sonnet 5.5 and Muse Wonder Full. It is now available on OpenRouter with a 1M-token context window, 2.8T total and 104B active parameters, priced at $2 per million input tokens and $10 per million output tokens, 33% lower than Kimi K3.
Reference imageEditorial responsibility
- No named human review is recorded for this page.
- Reports are grouped by semantic similarity and deterministic rules. Language models may assist titles, summaries, translation and cross-source analysis; the page reads the event directly, while its address stays stable when the title changes.
- Summary covers the current reports
Cross-source coverage
Reporting timeline
Moonshot Releases Kimi K3.1 Pro, Reportedly Beating Claude Opus 5.5 on Benchmarks
Moonshot has released Kimi K3.1 Pro, according to a post from ChrisGPT on X. The model reportedly outperforms Claude Opus 5.5 on SWE Marathon, Humanity's Last Exam, and several other benchmarks. The poster notes that they had disclosed early that Kimi K3.1 Pro was coming and would be cheaper than Kimi K3. The post describes the release as impressive, highlighting that Moonshot is continuing to push the frontier while driving down inference costs. The benchmark claims and pricing comparison are attributed to the poster and have not been independently verified in the source item.
Read sourceKimi K3.1 Pro Launches on OpenRouter With 1M Context and Multimodal Support
Moonshot's Kimi K3.1 Pro is now available on OpenRouter, according to a post from the OpenRouter account. The model is described as a 1M context, multimodal open-weight model that was post-trained from Kimi K3 for long-horizon coding and agentic work. Pricing is listed at $2 per million input tokens and $10 per million output tokens. The post includes a link for users to try the model. No further details on benchmarks, availability regions, or licensing terms were provided in the source item.
Read sourceKimi K3.1 Pro Debuts 4th on Intelligence Index, Set to Lead Open-Weights Models
Moonshot's Kimi K3.1 Pro launched at 4th place on the Artificial Analysis Intelligence Index with a score of 55, an 11-point gain over Kimi K3 (max, 44). It trails only Claude Opus 5.5 (58), Claude Sonnet 5.5 (56), and Muse Wonder Full (56). Once weights are published, it will be the highest-scoring open-weights model, narrowing the open-versus-proprietary gap from 12 points to 3. Agentic performance improved sharply: 3rd on GDPval-AA with 1,846 Elo (up 312), highest on Terminal-Bench 4.0 at 67% (up from 13%), and AutomationBench-AA rising from 58% to 73%. Other evals gained moderately: Humanity's Last Exam 47% to 52%, CritPt 23% to 25%, SciCode 59% to 61%, and AA-Omniscience Index 20 to 30. Cost per Task is roughly equal at $1.98. The model has a 1M-token context window, 2.8T total and 104B active parameters, text and image input, and pricing of $2/$10/$0.20 per 1M input/output/cache hit tokens, 33% lower than Kimi K3.
Read sourceShow 4 older updatesHide older updates
Kimi K3.1 Pro Beats Opus 5.5 on Several Benchmarks at Lower Price
A post reports that Kimi K3.1 Pro outperforms Opus 5.5 on several benchmarks, including SWE Marathon (51.7 vs 46.3), OfficeQA Pro (64.4 vs 62.2), Humanity's Last Exam (70.0 vs 67.7), GDP.pdf (34.6 vs 30.6), and ZeroBench (63 vs 61). The post states Opus 5.5 remains better overall but calls Kimi K3.1 Pro a good update, adding that it also competes with GPT-6 Astra on several benchmarks, beating it on some while Astra leads on Terminal-Bench Science and Kimi Code Bench Long-horizon. Pricing is highlighted: Kimi K3.1 Pro at $2/M input and $10/M output, versus Kimi K3 at $3/$15, Opus 5.5 at $4/$20, Sonnet 5.5 and GPT-6.1 Sol at $2/$10, and GPT-6 Astra at $10/$50. The post says Kimi K3.1 Pro has 2.8T total parameters and a 1M-token context window, calling it one of the strongest open-weight models available.
Kimi K3.1 Pro Tops US Frontier Models on 6 of 14 Benchmarks at Lower Token Prices
A post on X claims that Moonshot's Kimi K3.1 Pro outperforms every listed US frontier model on 6 of 14 benchmarks in Moonshot's evaluations. The post states the model's standard token rates are $2 per million input tokens and $10 per million output tokens, which it says are 50% below Opus 5.5 at $4/$20 and 80% below GPT-6 Astra at $10/$50. The author characterizes the release as unexpected and describes the offering as state-of-the-art at a much cheaper price, also noting local running with enough compute. The benchmark and pricing figures are presented as the poster's claims based on Moonshot's own evaluations, not independently verified results.
Read sourceKimi K3.1 Leaked as Coming Tomorrow or Monday, Cheaper Than K3
A leak claims Moonshot's Kimi K3.1 will launch tomorrow or Monday, following an official teaser. The model reportedly appears on LMArena under the codename "Kibitz," where the poster says they tested it. According to the post, Kimi K3.1 performs well at 3D, color selection, water reflection rendering, and composition, though it feels somewhat slow. It is said to be cheaper than Kimi K3 with better token efficiency, described as a killer model. The poster rates it comparable to Fable and Astra but still behind them, while noting Kimi is catching up fast, and calls it potentially one of the best open-source models to come.
Read sourceKimi K3.1 Expected Tomorrow, Possibly Monday, Cheaper and More Token-Efficient Than K3
A post on X states that Kimi K3.1 is expected tomorrow, possibly Monday, and claims it is cheaper than Kimi K3 while being more token-efficient. The author says they have used it frequently over the past week and that it performs well at Three.js, with a linked demo testing its reverse-engineering capabilities.