Kimi K3.1 Pro Open Weights Scheduled for November 4 — 2.8T MoE Follows K3 Open Pattern
Moonshot AI will release the full weights for Kimi K3.1 Pro on November 4, 2026, following the same API-first-then-open pattern it established with Kimi K2.6 and Kimi K3.
The Release
K3.1 Pro weights go public approximately 12 days after the model’s late-October API launch. The timeline mirrors Moonshot’s established practice: Kimi K3 debuted as a paid API in July 2026 and shipped an MIT-licensed weight release within weeks.
K3.1 Pro is a 2.8T parameter mixture-of-experts model with 104B active parameters per forward pass. API pricing before the open release is $2 per million input tokens and $10 per million output tokens — 33% cheaper on both input and output than Kimi K3’s launch pricing of $3/$15.
Benchmark Position
Terminal-Bench 4.0 is K3.1 Pro’s clearest reference point. Moonshot reported 65.7% at launch, placing the model above GPT-6 Astra (57.9%) and Claude Fable 5.1 (55.8%) but below Claude Sonnet 5.5 (70.6%).
On the Artificial Analysis Intelligence Index v5.0, K3.1 Pro scores 46 — behind the top frontier tier (Claude Opus 5.5 at 58, Gemini 4 Carbon at 59) but within reach of models priced significantly higher.
Context: A Crowded Open-Weight Moment
The November 4 weight release lands during one of the densest periods for open-weight releases in 2026. Mistral Large 4 went open-source on October 29, with Terminal-Bench 4.0 at 41% and BrowseComp at 84.5% — the strongest open-weight showing on BrowseComp to date. Ornith 1.5 holds the top open-weight slot on SWE-bench Verified at 86%.
K3.1 Pro’s Terminal-Bench score of 65.7% would place it above all current open-weight models on that benchmark, including Mistral Large 4 (41%) and Ornith 1.5 (no Terminal-Bench 4.0 entry). If the number holds under community evaluation, K3.1 Pro becomes the open-weight Terminal-Bench leader by a wide margin.
Caveats
The K3.1 Pro launch was complicated by reports of an alleged undisclosed influencer campaign on Chinese social media around October 25. Moonshot has not formally addressed the reports. The controversy does not affect the technical release on November 4, but community reception of the weights will be the cleaner measure of the model’s standing.
Additionally, Moonshot’s Terminal-Bench figure (65.7%) comes from the vendor’s own evaluation. Community reproduction on November 4 will determine whether the score holds under independent harness conditions.