PyyanThe AI Index

Pyyan / Language Models / Kimi K3.1 Pro

Kimi K3.1 Pro

Frontier coding numbers at a fifth of the price, with the weights promised for November.

currentMoonshot AI ↗Verified 2026-10-26

$2

per M input, against $10 out

Specifications

Context1Msource · as of 2026-10-26
Input ($/Mtok)$2source · as of 2026-10-26
Output ($/Mtok)$10source · as of 2026-10-26
LicenceClosed for now, weights promised 4 Nov 2026source · as of 2026-10-26 · medium confidence
Max output944Ksource · as of 2026-10-26 · medium confidence
Parameters3T class, exact count not publishedsource · as of 2026-10-26 · medium confidence
Released23 Oct 2026source · as of 2026-10-26
Model IDkimi-k3.1-prosource · as of 2026-10-26

Official links

Kimi K3.1 Pro model pageMoonshot AI official site

What matters

  • 65.7% on Terminal-Bench 4.0 in Moonshot's own table, against 66.4% for Claude Opus 5.5 and 58.2% for GPT-6 Astra
  • Claim-level hallucination rate down from 23.4% on K3 to 8.3%
  • Cyber training data deliberately removed, which Moonshot says costs about nine points on xdev-bench-hard
  • Full weights promised by 4 November 2026, so the licence is not yet knowable

Best for long-horizon coding when the budget is the constraint.

Compare with

All 5 side by side →

Kimi K3.1 Pro vs GPT-6 Astra →Kimi K3.1 Pro vs Claude Fable 5.1 →Kimi K3.1 Pro vs Claude Fable 5 →Kimi K3.1 Pro vs Gemini 4 Argon →Kimi K3.1 Pro vs Claude Opus 5.5 →

Also in Language Models

GPT-6 Astra · OpenAIClaude Fable 5.1 · AnthropicClaude Fable 5 · AnthropicGemini 4 Argon · GoogleClaude Opus 5.5 · Anthropic