Artificial AnalysisSource publication:
Loading content…
Moonshot released the reasoning model Kimi K3.1 Pro, which scores 55 on the Artificial Analysis Intelligence Index and places fourth, 11 points above Kimi K3's maximum of 44. Its largest gains are agentic: 1,846 Elo on GDPval-AA, 312 Elo above Kimi K3, and the highest score of any model on Terminal-Bench 4.0. Its weights are not yet published; Moonshot is expected to release them soon after launch, at which point it would be the highest-scoring open-weights model and the gap to proprietary models on the index would narrow from 12 points to 3.
Moonshot's new reasoning model Kimi K3.1 Pro scores 55 on the Artificial Analysis Intelligence Index, placing fourth and 11 points above Kimi K3's maximum of 44. Kimi K3 previously topped out at 44 on the index, leaving a clear distance to the leaders. On the Intelligence Index it trails only Claude Opus 5.5 (max with fallback, 58), Claude Sonnet 5.5 (max with fallback, 56) and Muse Wonder Full (max, 56).
The gains concentrate in agentic tasks: Kimi K3.1 Pro ranks third on GDPval-AA with 1,846 Elo and takes the highest score of any model on Terminal-Bench 4.0. Kimi K3 sits 312 Elo lower on GDPval-AA, and on Terminal-Bench 4.0 Kimi K3 scores 13% against 67% for Kimi K3.1 Pro. On GDPval-AA it trails only Claude Opus 5.5 (1,866 Elo) and Muse Wonder Full (1,857 Elo); on Terminal-Bench 4.0 it leads Claude Sonnet 5.5 (64%); AutomationBench-AA, which tests agentic SaaS workflows, rises from 58% to 73%.
Other evaluations move more moderately: Humanity's Last Exam rises from 47% to 52%, CritPt from 23% to 25% and SciCode from 59% to 61%. These are gains of a few percentage points over Kimi K3. The AA-Omniscience Index rises from 20 to 30, with the hallucination rate falling from 53% to 44% and accuracy rising from 48% to 51%.
Cost per Task is roughly equal to Kimi K3 at $1.98 per Task. Kimi K3.1 Pro uses 55K output tokens per Intelligence Index task, 14% more than Kimi K3 at 48K, while its first-party token prices are 33% lower. Input, output and cache-hit prices are $2, $10 and $0.20 per 1M tokens, each 33% lower than Kimi K3.
The report covers scores on the Artificial Analysis Intelligence Index and the named evaluations GDPval-AA, Terminal-Bench 4.0, AutomationBench-AA, Humanity's Last Exam, CritPt, SciCode and AA-Omniscience, plus a 1M-token context window, 2.8T total and 104B active parameters, text and image input, and pricing; the weights are not yet published, and Moonshot is expected to release them soon after launch.
The weights are not yet published, so the open-weights positioning rests on Moonshot's expectation of releasing them soon after launch; the scores are those reported at launch, and gains on other evaluations are clearly smaller than on the agentic ones.
