Source time Edition date
Close readSimon Willison
Introducing Kimi K3.1 Pro
介绍 Kimi K3.1 Pro
Kimi K3.1 Pro joins the crowded frontier-model race, so the decision for teams is whether to benchmark it against their current stack before committing budget or workflows to it.
Judgment
For platform teams weighing a switch off Claude Opus 5.5, Kimi K3.1 Pro offers half the output price against a 3-point Artificial Analysis deficit (55 vs 58) — but the open weights that make it strategically interesting do not exist yet, only a promise.
Basis
Simon Willison's link post reports Kimi K3.1 Pro went live on the Kimi API on 23rd October 2026, priced at $2.00 per million input tokens and $10.00 per million output — the same as GPT-6.1 Sol and half the price of Claude Opus 5.5. Artificial Analysis scored it 55, ahead of Claude Fable 5.1, GPT-6 Astra and Gemini 4 Argon (all 53) and behind only Claude Opus 5.5 (58); Willison notes this is the first time a soon-to-be open weight model has scored this highly against the frontier. Moonshot's own benchmarks put it approximately level with Opus 5.5 and GPT-6 Astra. A "Highspeed" tier runs 2-3x faster at twice the price, except input token cache hits, which are priced the same. The mechanism for buyers: an OpenAI-compatible endpoint means a low-cost trial today, while the weight release by Nov 4, 2026 is what would move evaluation from API cost to hosting cost.
- Claude Opus 5.558
- Kimi K3.1 Pro55
- Claude Fable 5.153
- GPT-6 Astra53
Unclear The source gives no license terms, parameter count, or hardware requirements for the promised open weights, and no absolute latency numbers for Highspeed beyond the 2-3x multiplier.
Next Check Moonshot's release channels on Nov 4, 2026 to confirm the weights actually ship and under what license, since that date is a stated promise rather than a delivered artifact.