Pyyan / Open-Weight Models / Kimi K3.1 Pro

Kimi K3.1 Pro

Frontier coding numbers at a fifth of the price, and the weights are now out.

currentMoonshot AI ↗Verified 2026-11-11

$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
LicenceKimi K3.1-Pro License (custom, conditions above $20M)source · as of 2026-11-11
Max output944Ksource · as of 2026-10-26 · medium confidence
Parameters2.8T total, 104B active (HF card says 3T class)source · as of 2026-11-11 · medium confidence
Released23 Oct 2026, weights 4 Novsource · as of 2026-11-11
Model IDkimi-k3.1-prosource · as of 2026-10-26

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
  • Weights published on 4 November 2026 as promised, about 1.56 TB, under Moonshot's own licence rather than a standard one
  • A Model as a Service business above $20M in revenue needs a separate agreement, and the Kimi name must be shown on products above 100M monthly users or $20M monthly revenue. Internal use and Moonshot's certified partners are exempt

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