Moonshot AI: Kimi K3.1 Pro Performance Data

Moonshot AI has released Kimi K3.1 Pro, a new reasoning model that has entered the fourth position on the Artificial Analysis Intelligence Index with a score of 55 points.

This score represents an 11-point increase over its predecessor, Kimi K3. The model is now positioned three points behind the highest-scoring proprietary systems, which include Claude Opus 5.5 at 58 points, Claude Sonnet 5.5 at 56 points, and Muse Wonder Full at 56 points.

Moonshot AI has announced that it plans to release the model weights for K3.1 Pro by November 4, 2026. If this release proceeds as scheduled, the model would become the highest-scoring open-weights model available, narrowing the performance difference between open and proprietary models to three points.

The new model is available through Kimi.ai, Kimi Work, Kimi Code, and the Kimi API. Moonshot AI is also providing a variant, K3.1 Pro Highspeed, which the company reports runs two to three times faster than the standard version.

The model's development involved significant post-training compute, resulting in measurable improvements across specific tasks. On the GDPval-AA benchmark, K3.1 Pro achieved an Elo of 1,846, a gain of 312 points over the previous version. On the Terminal-Bench 4.0 test, the model recorded 65.7 percent, compared to 13 percent for Kimi K3.

Moonshot AI reports that claim-level hallucination rates on the Kimi Hallu Bench Hard stress test declined from 23.4 percent to 8.3 percent. On internal financial benchmarks, the model gained 7.2 points in Financial Modeling and 11.6 points in Financial Research.

For 3D modeling and coding, K3.1 Pro scores 92.7 percent on the BenchCAD benchmark when using tools, which is close to the 96.2 percent recorded by Claude Opus 5.5. The model is capable of writing and running scripts for Blender and CAD software without MCP connectors.

The model maintains the same dimensions as Kimi K3, with 2.8 trillion total parameters and 104 billion active ones, while operating over a context window of one million tokens. API prices for the model have been reduced by 33 percent, with input costs set at $2 per million tokens and output costs at $10 per million tokens.

Moonshot AI has also addressed cybersecurity capabilities in the model's training. The company removed cyber-related data from the training phases for the open-weights release. On the internal xdev-bench-hard test, the uncleaned version solves 90.5 percent of tasks, while the version released after data cleaning solves 81.0 percent.