AI BriefingMoonshot AI launches Kimi K2.6 in Kimi Chat and API
Key point
Moonshot AI has released Kimi K2.6 via Chat and API, touting it as open-source state-of-the-art.
Details
Moonshot AI unveiled Kimi K2.6, touting it as open-source state-of-the-art for code and agentic workloads. The model is available via chat and agent mode on kimi.com, weights released on Hugging Face, and the platform.moonshot.ai API.
The model selector has 4 variants.
- K2.6 Instant: for fast responses
- K2.6 Thinking: for deeper reasoning
- K2.6 Agent: for research, slides, websites, docs, sheets
- K2.6 Agent Swarm: for large-scale search, long-form output, batch tasks
In benchmarks, it claimed Humanity's Last Exam with tools 54.0, SWE-Bench Pro 58.6, SWE-bench Multilingual 76.7, BrowseComp 83.2, Toolathlon 50.0, Charxiv with Python 86.7, and Math Vision with Python 93.2. In the comparison chart, it was placed alongside GPT-5.4 xhigh, Claude Opus 4.6 max effort, and Gemini 3.1 Pro thinking high, with particular emphasis on Kimi leading in SWE-bench Multilingual and BrowseComp.
The launch comes about 1 week after K2.6 Code Preview, which entered beta on April 13. Following K2.5 hybrid reasoning from earlier this year, Moonshot AI is now positioning open weights and aggressive agent scaling as key differentiators in the frontier race.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
Jevper Released: A Jev Classification Wrapper for OpenAI-Compatible Models with Automatic Logprobs Detection