Moonshot AI
Moonshot Kimi K2.5 upgraded MoE model with strong agentic capabilities.
Moonshot AI's 1T-parameter MoE instruct model — top-tier agentic and coding performance rivalling Claude Sonnet.
Kimi K2.5 high reasoning — mini-SWE-agent v2
Moonshotai-Kimi-K2-Instruct FC — rank #11
Kimi K2.5 — rank #2, max_steps=100, Moonshot AI Jan 2026
Kimi K2 Instruct — rank #55
Moonshot AI's September 2025 Kimi K2 instruct checkpoint — updated 1T MoE model with improved agentic and coding capabilities.
Kimi K2.5 high reasoning — mini-SWE-agent v2
Moonshot AI's 1T-parameter MoE reasoning model — extended chain-of-thought variant of Kimi-K2 for hard math and science problems.
Kimi K2 Thinking — rank #28
