Leaderboard / Kimi K3.1 Pro

Kimi K3.1 Pro Max

Scores

Buzz

• CAN AI BUILD BETTER AI? Of the 27 models tested on @ValsAI's AI research eval, only 7 were able to improve the trained model they were given on a fixed compute budget and none of them were open models. Claude Fable 5.2 had the highest score at 37.88, while the best open model w…@pratikg · 17 likes
• sokobanteg v0.2 live gpt-6.1 sol, gpt-6.1 terra, gpt-6 astra claude opus 5.5, claude sonnet 5.5, claude haiku 5.5, claude fable 5.2 gemini 4 argon, gemini 4 carbon qwen3.8 max 0902, kimi k3.1 pro, muse wonder, muse ember, grok 4.7 build https://t.co/tv7RNPSEVT@banteg · 15 likes
• Cybersecurity agents need the full picture: code, crash logs and screenshots. Now on Infron: • Kimi K3.1 Pro Cyber for vulnerability research. • HY4 Uncensored for visual triage + remediation. 1M context + tool calling on both. https://t.co/m3vBH42f1Z https://t.co/giVcR05gvh h…@InfronAI · 9 likes

Evidence

Coding 62.0 ±4.1

Agents 62.7 ±1.2

Reasoning 63.5 ±2.7

Math 62.5 ±1.6

Knowledge 60.5 ±2.7

Frontend 60.7 ±0.7

General 60.2 ±4.1

Results not used in the score

Names matched to this model

kimi-k3-1-pro-max

Compare Kimi K3.1 Pro Max with other models