Moonshot AI Unleashes Kimi K3.1 Pro, Narrowing the Gap on Top AI Models

Chinese AI firm Moonshot AI released Kimi K3.1 Pro on October 23, 2026, a new reasoning model designed to push the boundaries of post-training compute. The release marks a significant step for the company as it seeks to compete with top proprietary systems.

The model's performance on the Artificial Analysis Intelligence Index has jumped to 55 points, an 11-point increase over its predecessor, Kimi K3. This score places K3.1 Pro at number four, just three points behind the leading proprietary models, including Anthropic's Claude Opus 5.5. The gap between open-weight and proprietary models is on track to shrink from 12 points to just 3.

Moonshot AI plans to release the model weights by November 4, 2026. If this schedule is met, K3.1 Pro will become the highest-scoring open-weights model available. The company is also shipping a faster variant called K3.1 Pro Highspeed, which it claims runs two to three times faster.

The model demonstrates its strongest gains in agentic tasks. On the GDPval-AA benchmark, K3.1 Pro achieves an Elo score of 1,846, placing it just behind Claude Opus 5.5 and Muse Wonder Full. Its performance on the Terminal-Bench 4.0 benchmark reached 65.7 percent, a significant leap from the 13 percent score of K3.

Coding and 3D modeling capabilities have also received a major boost. The model can interface with game engines like Unity and Unreal Engine to handle fixes or full project builds. On the BenchCAD benchmark, it scores 92.7 percent with tools, nearing the performance of Claude Opus 5.5 and GPT-6 Astra. Moonshot states the model is strongest at parametric modeling of mechanical parts and architectural models.

Despite these enhancements, the model's architecture remains unchanged, with 2.8 trillion total parameters and 104 billion active ones. It supports a context window of one million tokens and can process both text and images. The model sits at the top of the Pareto frontier, meaning no model with a comparable parameter count scores higher on the Intelligence Index.

The API price for K3.1 Pro has dropped by 33 percent. Input costs are now $2 per million tokens, down from $3, while output costs are $10, down from $15. Moonshot reports that the cost for each Intelligence Index task is $1.98, nearly identical to K3's $2.00.

Moonshot also addressed cybersecurity concerns in the model's development. The company stripped cyber-related data from the training phases for the open-weights release. On an internal benchmark that reproduces recently disclosed vulnerabilities, the cleaned model solves 81.0 percent of tasks, while the uncleaned version solves 90.5 percent.

Moonshot AI states that no open-weights model should push the absolute frontier of offensive cyber capabilities. However, the company aims to raise the capabilities of open models for defensive purposes, such as auditing legacy systems. The availability of K3.1 Pro through Kimi.ai, Kimi Work, Kimi Code, and the Kimi API marks a key moment for the company's strategy.