Settled
Moonshot unveils Kimi K3.1 Pro, a 3T-class open-weight model, with cyber data stripped from training
Moonshot's Kimi K3.1 Pro has a 1M-token context window, and the company says full weights arrive by Nov. 4. It lists 65.7% on Terminal-Bench 4.0 versus 66.4% for Claude Opus 5.5. Moonshot says it removed cyber-related data from training, and its internal xdev-bench-hard score fell to 81.0% from 90.5%. Benchmarks are Moonshot's own. License and pricing are not stated.
- HEAT
- OUTLETS
- 7
- FIRST SEEN
- LAST SIGNAL
Primary source
Kimikimi.ai/blog/kimi-k3-1-proCoverage · 7 articles, oldest first
Friday, October 23, 2026
Artificial Analysis
artificialanalysis.ai/models/kimi-k3-1-pro
Saturday, October 24, 2026
Latent Space
latent.space/p/ainews-kimi-31/