Kimi K3.1 Pro debuts 4th on Intelligence Index with big agentic gains, prices cut 33%
ArtificialAnlys · x · 2026-10-24
Artificial Analysis evaluated Moonshot's Kimi K3.1 Pro: it scores 55 on the Intelligence Index, up 11 points from Kimi K3 (44), ranking 4th behind only Claude Opus 5.5 (58), Claude Sonnet 5.5 (56), and Muse Wonder Full (56). Once weights drop, it will be the top open-weights model, narrowing the open-vs-proprietary gap from 12 to 3 points.
Key results
- Big agentic gains: 3rd on GDPval-AA with 1,846 Elo (+312 vs K3); highest-ever score on Terminal-Bench 4.0 at 67% (vs 13% for K3), ahead of Claude Sonnet 5.5 (64%); AutomationBench-AA rises from 58% to 73%.
- Moderate gains elsewhere: Humanity's Last Exam 47%→52%, CritPt 23%→25%, SciCode 59%→61%; AA-Omniscience 20→30 as hallucination rate falls 53%→44% and accuracy rises to 51%.
- Pricing: token prices cut 33% ($2/$10/$0.20 per 1M input/output/cache-hit tokens); cost per task roughly flat at $1.98 since output tokens grew 14% (55K vs 48K).
- Specs: 2.8T total / 104B active parameters, 1M context, text+image input; weights expected to be released soon, which would make it the highest-scoring open-weights model.
More from Models
- Community claims Kimi k3.1 is the most 'benchmaxxed' model ever, alleging distillation — imjustnewatai · 2026-10-24
- Leaker: Grok 4.8 arriving next week — mark_k · 2026-10-24
- Labs are far ahead of what they release — and the gap may be widening, speculates X user — MoonL88537 · 2026-10-24
- API Pricing Cut from $15 to $10 per Million Tokens — cedric_chee · 2026-10-24
- MiniMax's 430B model beats 3T-scale rivals on AA knowledge benchmark — teortaxesTex · 2026-10-24
- Hand-crafted regex pretokenizers quietly do heavy lifting in every frontier LLM — YouJiacheng · 2026-10-24