Kimi K3.1 Pro and Highspeed models
Archived — this story has rotated out of today’s deck. It is kept here in full.
The gist
Chinese AI company Moonshot AI released Kimi K3.1 Pro, a 3T-class open-weight model with a 1-million-token context window.
Its planned weight release could narrow the gap between open and proprietary models.
Background
Moonshot AI is a Chinese AI company behind the Kimi product family, which includes Kimi.ai, Kimi Work, Kimi Code and the Kimi API. K3.1 Pro is a post-trained version of its prior model Kimi K3, trained with substantially more post-training compute and reinforcement learning across hundreds of downstream domains. The launch is notable because the model is described as open-weight, with full weights due by November 4, 2026, and because an independent evaluator, Artificial Analysis, scores it close to leading proprietary models.
How it unfolded
- Oct 23, 2026Moonshot AI announced Kimi K3.1 Pro and K3.1 Pro Highspeed across Kimi.ai, its mobile apps, Kimi Work, Kimi Code and the Kimi API. Highspeed was stated to deliver 2-3x the speed of K3.1 Pro with identical capabilities.
- Oct 23, 2026Artificial Analysis listed an Intelligence Index score of 55 for K3.1 Pro, behind Claude Opus 5.5 (58), Claude Sonnet 5.5 (56) and Muse Wonder Full (56).
- Oct 24, 2026Kimi officially released K3.1 Pro, described as its most powerful 3-trillion-parameter model to date; community quota was fully depleted in 10 minutes. Kimi also launched the Hosted Agents managed agent service.
- Nov 4, 2026Full model weights of K3.1 Pro are due to be released by this date.
Who’s saying what
- Official
- Moonshot AI says K3.1 Pro is its most capable 3T-class open-weight model to date, citing gains in coding, 3D creation, financial research and agentic capabilities, and says Highspeed offers 2-3x the speed for more responsive interactions.
- Analysts
- Kingy.ai notes the speed claim applies to model output, so if most of a job's time goes to running tools, faster token generation improves only part of total duration.
- Analysts
- The Decoder reports that if Moonshot releases the weights as planned by November 4, 2026, K3.1 Pro would become the highest-scoring open-weights model available, shrinking the open-proprietary gap from 12 to 3 points.
Still unverified
Kimi's benchmark comparisons are vendor-reported results, not independent measurements. Kingy.ai notes the '3T-class' description is rounded and should not be turned into a claim that the model has exactly three trillion parameters, while ForkLog reports 2.8 trillion parameters with 104 billion activated per query, differing from the 3T-class description.