Moonshot says that on Terminal-Bench 4.0, its own benchmark run with the Claude-Code harness, Kimi K3.1 Pro scores 65.7%, ahead of GPT-6 Astra (58.2%) and narrowly behind Claude Opus 5.5 (66.4%); the other models scores are the maximum values reported in their own blog posts or on the official leaderboard.

The exact words it rests on

K3.1 Pro scores 65.7%, ahead of GPT-6 Astra (58.2%) and narrowly behind Claude Opus 5.5 (66.4%).

What the source said when we opened it, on 24 Oct 2026.

The source

Introducing Kimi K3.1 Pro
Moonshot AI (Kimi) · 2026-10-23

Open the sourcearchived copy

Checked

Checked by the notis newsroom on 24 Oct 2026, against the source above.

The number in it

• 65.7% · Kimi K3.1 Pro, benchmark score (Terminal-Bench 4.0)
• 66.4% · Claude Opus 5.5, benchmark score (Terminal-Bench 4.0 (as reported by Moonshot))

In the story

Moonshot releases Kimi K3.1 Pro with a 1M-token window and says weights follow by 4 November 24 Oct 2026