Kimi releases K3.1 Pro and says open weights will follow by 4 November

Kimi's technical post introduces Kimi K3.1 Pro as the first model in the K3.1 series and its most capable 3T-class open-weight model to date, with a 1-million-token context window. It also releases Kimi K3.1 Pro Highspeed, which it says delivers 2-3× the speed of K3.1 Pro. Both are available on Kimi.ai, the Kimi mobile apps, Kimi Work, Kimi Code and the Kimi API, under the API names kimi-k3.1-pro and kimi-k3.1-pro-highspeed. Kimi says Zero Data Retention is available for API customers, and that Kimi Enterprise access is off by default at launch. Full model weights will be released by 4 November 2026.

The blog page does not print a publication date. Its page data records a modified date of 24 October 2026. Moonshot's forum post announcing the same model is timestamped 23 October 2026, 20:12 UTC, which is the date used here.

On Terminal-Bench 4.0, Kimi says K3.1 Pro scores 65.7%, ahead of GPT-6 Astra at 58.2% and narrowly behind Claude Opus 5.5 at 66.4%. Those comparisons are Kimi's. For the open-weight release, Kimi says it removed cyber-related data from mid-training and from reinforcement-learning environments. On its in-house xdev-bench-hard, it says a checkpoint without those removals solves 90.5% of problems at pass@3, and the final K3.1 Pro reaches 81.0%. It says that cyber capability remains behind frontier proprietary models that have been publicly available for at least three months.

What it means for business: the model is on Kimi's API now; the weights are not out yet. The 4 November date is Kimi's own deadline for the weights. Treat the benchmark scores as the company's figures. Check sector and procurement rules before sending business data to a Chinese-developed API, and read the API's data terms, including the Zero Data Retention option.