Gate News message, April 20 — Moonshot AI has released Kimi K2.6, now available in both chat and Agent modes on kimi.com. The model achieved leading scores across multiple open-source coding and tool-use benchmarks, including HLE with tools, SWE-Bench Pro, SWE-bench Multilingual, and BrowseComp.
Kimi K2.6 supports over 4,000 tool invocations and can run continuously for more than 12 hours, with multilingual capabilities for Rust, Go, and Python. The model’s parallel Agent functionality has been upgraded to handle up to 300 concurrent sub-agents per run and 4,000 steps, now deployed in autonomous scenarios such as OpenClaw and Hermes Agent.
Related News
TokenAI and DeBox Partner on AI-Powered Web3 Social
Canva announces deep integration with Claude, enabling the transformation of AI drafts into finished designs
NVIDIA announces the world’s first “Ising model” built for quantum AI