Moonshot announced that the Kimi K3.1 Pro model is now accessible via its API. The company has not yet published the full model weights, but it has committed to releasing the complete weights by November 4, 2026.
Pricing and variants
- US dollar pricing: $2.00 per million input tokens and $10.00 per million output tokens. Moonshot notes this matches GPT‑6.1 Sol’s pricing and is roughly half the price of Claude Opus 5.5.
- A “Highspeed” tier is offered that Moonshot says is 2–3× faster; this tier is generally priced at twice the normal rate. An exception applies to input token cache hits, which are charged at the same rate as the standard tier.
Performance and benchmarks
According to Moonshot’s own benchmarks, Kimi K3.1 Pro performs approximately on par with Claude Opus 5.5 and GPT‑6 Astra—outperforming them on some tests and closely trailing on others.
The independent reviewer Artificial Analysis gave Kimi a score of 55. That places it ahead of Claude Fable 5.1, GPT‑6 Astra, and Google Gemini 4 Argon (all at 53), and behind only Claude Opus 5.5 (58). The article highlights that this is the first time a model that will soon become open‑weight has scored this highly relative to frontier closed models.
Hands‑on test via the API
The author tested Kimi K3.1 Pro through the Moonshot API using the llm openai endpoint command they added to LLM 0.32 in August. The test prompt requested an SVG graphic: “Generate an SVG of a pelican riding a bicycle.” The example shows how the llm openai endpoint can target any OpenAI‑compatible endpoint but does not automatically log responses, so stdout and stderr were redirected to files to capture the model outputs and reasoning traces.
Kimi K3.1 Pro produced outputs at three supported reasoning levels—low, medium, and max—and the author includes the max‑level result. For comparison, the author also generated fresh outputs from Kimi K3 on the same three reasoning settings.
Why this matters
Making Kimi K3.1 Pro available on the API while committing to release the model weights may be significant for developers and researchers who want access to a competitive model with more open availability. Based on the announced pricing and the reported benchmark results, Kimi K3.1 Pro appears to be a competitive offering among current frontier models, particularly for users weighing cost versus performance.
At the time of publication the weights had not yet been released; Moonshot set the public release date for the full weights as November 4, 2026.
