Moonshot AI released Kimi K3.1 Pro via its API with a commitment to open-weight availability by November 4, 2026, marking the first soon-to-be-open model to reach frontier-tier benchmark scores. Pricing is set at $2.00 per million input tokens and $10 per million output tokens, matching GPT-6.1 Sol and undercutting Claude Opus 5.5 by half; a "Highspeed" variant runs 2–3× faster at double the base price, though cached input tokens retain standard pricing. Moonshot’s internal benchmarks position K3.1 Pro roughly level with Opus 5.5 and GPT-6 Astra, while Artificial Analysis rates it 55 — ahead of Claude Fable 5.1, GPT-6 Astra, and Gemini 4 Argon (all 53) and behind only Opus 5.5 at 58. The model supports three reasoning effort levels (low, medium, max). Simon Willison tested it through the LLM 0.32 `llm openai endpoint` command against the Moonshot API (`https://api.moonshot.ai/v1`), capturing reasoning traces via stdout/stderr redirection, and compared SVG pelican outputs across reasoning tiers against the prior Kimi K3 release.