Kimi-Dev

model paper
Moonshot AI
Open-source SWE LLM achieving 60.4% on SWE-bench Verified. Uses agentless training as skill prior for SWE-Agents.
Paper (arXiv)GitHubHuggingFace

Outputs 2

Kimi-Dev-72B

model
Parameters 72B

Kimi-Dev Paper

paper
arXiv HTML
codingagentic

Notes

Model date approximate.