Source: Kimi tech blog (Moonshot)
Moonshot releases Kimi K3.1 Pro in its apps and API, with open weights promised by Nov 4
Why it matters
Kimi K3.1 Pro (2.8T total / 104B active parameters, 1M-token context, text and image input) is live on Kimi.ai, Kimi Work, Kimi Code and the Kimi API as kimi-k3.1-pro, plus kimi-k3.1-pro-highspeed, which Moonshot says is 2-3x faster; Moonshot says full weights will be released by Nov 4. API pricing is $2 input / $10 output / $0.20 cached input per million tokens, down from Kimi K3's $3 / $15 / $0.30.
Artificial Analysis scores it 55 on its Intelligence Index, 4th at launch, and 67% on Terminal-Bench 4.0 versus 13% for Kimi K3; Moonshot's own Terminal-Bench run reports 65.7%. Cheaper tokens don't mean cheaper tasks: Artificial Analysis says it uses more output tokens per task than K3 (55K vs 48K), and its model pages list about the same cost per Intelligence Index task ($2.11 vs $2.00 for K3). Budget by task, not by token price, and wait for the published weights before planning self-hosting.