Kimi K3.1 Pro

Not yet independently verified. Moonshot's blog prints no release date, so the 23 October release rests on its pricing page banner and Artificial Analysis; the benchmark results are Moonshot's own. The weights were published on Hugging Face on 2 November 2026 under Moonshot's own Kimi K3.1 Pro License. We will update this when it can be confirmed, and remove this note.

Kimi K3.1 Pro is a large language model from the Beijing company Moonshot AI, the first in its K3.1 series, released on 23 October 2026 on Kimi.ai, Kimi Work, Kimi Code and the Kimi API. Moonshot describes it as a post-trained version of Kimi K3 with much more reinforcement-learning compute, in the 3-trillion-parameter class, with a context window of 1,048,576 tokens. A second version, Kimi K3.1 Pro Highspeed, runs two to three times faster at twice the price.

Update, 10 November 2026: Moonshot has published the weights. The Hugging Face repository moonshotai/Kimi-K3.1-Pro was created on 2 November 2026 and last modified on 5 November, and its model card names the Kimi K3.1 Pro License. The licence text, copyright 2026 Moonshot AI, grants use, modification and redistribution with two conditions. A business running a "Model as a Service" product with more than 20 million US dollars in revenue over 12 months must agree terms with Moonshot AI first. Products with more than 100 million monthly users or more than 20 million dollars in monthly revenue must show "Kimi K3.1-Pro" in the interface. The licence excludes internal use and use through Moonshot's official products or certified partners. The benchmark results remain Moonshot's own.

What it does

Moonshot pitches K3.1 Pro at coding agents, financial research, 3D creation and other multi-step agentic work. It reports 65.7% on Terminal-Bench 4.0, a test of coding agents working on constrained terminal tasks, and 61.2% on its own long-horizon coding benchmark against 55.95% for K3; those are Moonshot's evaluations, so no score is stored here. Artificial Analysis places it fourth of 215 models on its Intelligence Index, with a score of 55, and measured about 80.7 output tokens per second.

The API price is $2.00 per million input tokens and $10.00 per million output tokens, with cached input at $0.20; Highspeed costs $4.00 and $20.00. Writing to the cache costs $2.00 per million tokens for five minutes and $4.00 for an hour on the standard model. On the same pricing page Moonshot says the K2 series closes to new users on 31 October 2026 and will be retired by the end of December, so teams still building on K2 have a migration to plan.

Who it is for

The model suits teams that want a long-context coding and agent model at a lower list price than the largest proprietary models, and that may later want to run the weights themselves. Moonshot has since published the weights on Hugging Face under its own Kimi K3.1 Pro License, described in the update above; Artificial Analysis labelled the model proprietary at launch. Enterprise administrators must switch it on for their workspace, as it is off by default.

Key facts

Maker: Moonshot AI Released: 23 October 2026 Context: 1,048,576 tokens Price: $2.00 input, $10.00 output, $0.20 cached input per 1M tokens (Highspeed $4.00 and $20.00) Weights: published on Hugging Face on 2 November 2026 under the Kimi K3.1 Pro License Predecessor: Kimi K3; K2 series retired by end of December 2026