Kimi K3.1 Pro

Kimi K3.1 Pro is a large language model from Moonshot AI (Kimi). AI Stupid Level re-tests it on a fixed schedule (coding every four hours, reasoning and tool use daily, with an hourly canary probe) and publishes a Combined score from 0 to 100: half coding, a quarter reasoning and a quarter tool use, where higher is better. Kimi K3.1 Pro is 19th of the 24 models on the Combined board in the last 24 hours, scoring 81 ± 2.0 out of 100, and its trend is holding steady. On the individual boards it is joint 22nd of the 24 models on the Coding board. "Joint" means its lead over, or gap to, the models around it is smaller than the measurement can resolve, so they share the place. The scores come from our own runs through the provider's public API, not from vendor-reported numbers, so they show how the model behaves as it is served.

As of 26 Oct 2026, 11:37 UTC (24-hour boards) · tracked since 26 Oct 2026 (0 days) · last re-tested 26 Oct 2026, 11:32 UTC.

On the same schedule as every model on the board: coding every four hours, reasoning and tool use daily, with an hourly canary probe. The coding suite runs every four hours: seven tasks, seven attempts each, most of them real debugging jobs in which the model gets a small working project and a bug report and has to find and fix the fault. It is graded by running the project's own tests, including tests the model never sees, and scored on nine measures led by correctness (60%), with stability, edge cases and debugging at 10% each; complexity and efficiency are scored but carry no weight. The reasoning suite runs daily: four long working sessions, scored on correctness, recovery after a failed step, and three continuity measures (memory retention, plan coherence and context use) checked by running code against rules stated earlier in the conversation. The tool-use suite runs daily: nine tasks in sandboxed machines, scored on task completion (30%), tool selection (20%), parameter accuracy, efficiency, error handling, context awareness and safety. A suite that is missing or out of date drops out of the Combined score and the others are reweighted. Full details, including every weight, are on the benchmarking methodology page.

As of 26 Oct 2026, 11:05 UTC, no detector has fired. The coding detector is still in its first ten days of daily results on the current version of the tests and cannot fire yet. On the Combined board over the last 24 hours its trend is holding steady, at 81 ± 2.0 out of 100.

This is the question the platform exists to answer. A provider can change a model behind the same API name, and without continuous re-testing that change is invisible to the people relying on it. Each of Kimi K3.1 Pro's suites has its own Page-Hinkley (a CUSUM-family test) change-point detector, run on daily results, which separates a sustained decline from ordinary run-to-run noise; it arms after ten days of daily results on the current version of the tests. An hourly canary of two fixed probes watches for a sudden collapse, tests it against the previous week with Welch's t-test, and records an incident when the fall is large and significant. A Page-Hinkley detection is marked on the chart above and changes Kimi K3.1 Pro's status on the drift monitor. If you are wondering whether Kimi K3.1 Pro has been nerfed, or why it seems dumber lately, see the current answer for every Kimi model and how AI drift detection works for the method.

Closest to Kimi K3.1 Pro on the Combined board in the last 24 hours: GPT-6.1 Sol (81), GPT-5.5 (82) and Claude Opus 4.7 (78), each with a head-to-head page. Every model gets the same tasks on the same schedule, so places are comparable across vendors. Each pair page puts both models on the four boards with the leaderboard's statistical ties, list price per 1M tokens, cost per correct result over the last 30 days and drift state; the full boards are on the leaderboard.

Closest on the same tests

• Kimi K3.1 Pro vs GPT-6.1 Sol
• Kimi K3.1 Pro vs GPT-5.5
• Kimi K3.1 Pro vs Claude Opus 4.7

Against each vendor's best

• Kimi K3.1 Pro vs GPT-5.6 Luna
• Kimi K3.1 Pro vs Claude Fable 5.1
• Kimi K3.1 Pro vs Gemini 3.8 Flash
• Kimi K3.1 Pro vs DeepSeek V4 Pro
• Kimi K3.1 Pro vs GLM-5.3

Every Moonshot AI model on the board, every model family compared, the best LLM for coding, the best AI model for agents and Moonshot AI API status. Sign in and star Kimi K3.1 Pro to add it to your watchlist: it then appears at the top of the leaderboards with its place on each board, and we email you if its coding score falls five points or more in a week, plus a weekly summary of what changed.