Moonshot AI’s Kimi K3.1 Pro Narrows Gap With Top Western Rivals

The landscape of artificial intelligence leadership is shifting rapidly as open-weights models begin to challenge the dominance of closed, proprietary systems. Today, Moonshot AI announced the release of Kimi K3.1 Pro, a significant upgrade to its previous K3 model that has immediately captured attention in the tech community.

The new model has secured the fourth position on the Artificial Analysis Intelligence Index with a score of 55. This represents an 11-point improvement over its predecessor and places it just three points behind the current leaders. Only Claude Opus 5.5, Claude Sonnet 5.5, and Muse Wonder Full currently rank higher on this specific metric.

If Moonshot follows through on its plan to release the model weights by November 4, 2026, Kimi K3.1 Pro would become the highest-scoring open-weights model available. This move would effectively shrink the performance gap between open and proprietary models from 12 points to just three, assuming no other major releases occur in the interim.

The most dramatic improvements are evident in agentic tasks and coding. On the GDPval-AA benchmark, the new model reaches an Elo rating of 1,846, a 312-point leap from K3. In Terminal-Bench 4.0, Moonshot reports a success rate of 65.7 percent, surpassing GPT-6 Astra and trailing only slightly behind Claude Opus 5.5. Notably, the older K3 model managed only 13 percent on this same test.

Reliability has also seen a sharp uptick. Internal testing on Moonshot’s Kimi Hallu Bench Hard shows that claim-level hallucinations dropped from 23.4 percent to 8.3 percent. While the company notes this is a stress test rather than a general error rate measure, it remains higher than GPT-6 Astra’s 5.5 percent. The model also posts gains in financial modeling and research, and it tops all tested models on the OfficeQA Pro v2 benchmark for general office tasks.

Beyond coding, Kimi K3.1 Pro demonstrates strong capabilities in 3D modeling. It can interact directly with game engines such as Unity, Unreal Engine, and Godot. On the BenchCAD benchmark, the model scores 92.7 percent when using tools and 72.8 percent without them, figures that are nearly identical to Claude Opus 5.5. Moonshot has also launched a "3D Make" feature on its website, allowing users to preview, edit, and export 3D designs.

Pricing has been adjusted to reflect the model's efficiency. While the parameter count remains at 2.8 trillion total and 104 billion active, the cost per million input tokens has dropped by 33 percent to $2. Output costs are now $10 per million tokens, and cache hits cost $0.20. A Highspeed variant, which runs two to three times faster, is also available through Kimi.ai, Kimi Work, and the API.

Moonshot has also addressed cybersecurity concerns by stripping cyber-related data from the training corpus for the open-weights version. On an internal benchmark that reproduces recent software vulnerabilities, the uncleaned version solved 90.5 percent of tasks, while the cleaned K3.1 Pro solved 81.0 percent. This is still ahead of the previous K3 model, which solved 33.3 percent, but slightly behind DeepSeek-V4.1-Flash at 85.7 percent.

The company argues that open models should not push the absolute frontier of offensive cyber capabilities but should gradually raise the bar for defensive purposes, such as auditing legacy systems. This approach acknowledges that even lower-scoring open models can write sophisticated exploits, a point previously highlighted by Anthropic regarding GLM-5.3.

As these models become more accessible and affordable, the distinction between open and closed-source AI continues to blur. The rapid progress of Kimi K3.1 Pro suggests that the competition for the top spot in the intelligence index is far from over, with more updates and weight releases expected in the coming weeks.