Kimi unveils K3.1 Pro, a 3T-class open-weight model with a 1M-token context window, and K3.1 Pro Highspeed, with full weights to be released by November 4 (Kimi)
Kimi reports coding results close to Opus 5.5, with full weights still pending release.
On Terminal-Bench 4.0, Kimi reports K3.1 Pro scored 65.7%, versus 66.4% for Claude Opus 5.5 and 58.2% for GPT-6 Astra. The multimodal model uses a mixture-of-experts architecture with 2.8 trillion total parameters and 104 billion active parameters. Artificial Analysis places it fourth on its Intelligence Index and says it would lead open-weight models once the weights are published. Full weights are promised by November 4.
Techmeme's note
On Terminal-Bench 4.0, Kimi reports K3.1 Pro scored 65.7%, versus 66.4% for Claude Opus 5.5 and 58.2% for GPT-6 Astra. The multimodal model uses a mixture-of-experts architecture with 2.8 trillion total parameters and 104 billion active parameters. Artificial Analysis places it fourth on its Intelligence Index and says it would lead open-weight models once the weights are published. Full weights are promised by November 4.
Techmeme's note
score 8