Kimi K2.6: Moonshot AI's Open-Weight Model That Can Code for 13 Hours Straight
Published on 21.04.2026
Kimi K2.6: Moonshot AI's Open-Weight Model That Can Code for 13 Hours Straight
TLDR: Moonshot AI released Kimi K2.6, a fully open-weight model that scores 80.2% on SWE-Bench Verified and ran autonomously for 13 hours making over 1,000 tool calls in a real-world test. It's available now in the Kilo Code platform and supports scaling to 300 parallel sub-agents.
External Links (1)
Sign in to bookmark these links