55 Points: How Kimi K3.1 Pro Narrowed the Gap With Top AI Rivals

A new entry from China’s Moonshot AI has just pushed into the upper echelon of artificial intelligence performance. The company released Kimi K3.1 Pro on October 23, 2026, and the model immediately secured the fourth spot on the Artificial Analysis Intelligence Index with a score of 55. This result places it just three points behind the leading proprietary models, marking a significant shift in the competitive landscape for AI capabilities.

To understand how quickly this happened, it helps to look at the trajectory of the Kimi series. The new release is a post-trained version of the earlier Kimi K3, which scored 44 points on the same index. By applying significantly more post-training compute, Moonshot achieved an 11-point jump. Currently, only three models rank higher: Claude Opus 5.5 at 58, and both Claude Sonnet 5.5 and Muse Wonder Full at 56. If Moonshot follows through on its plan to release the model weights by November 4, 2026, K3.1 Pro would become the highest-scoring open-weights model available. This would shrink the performance gap between open and proprietary systems from 12 points to just three.

The most dramatic improvements are visible in agentic tasks, where the model demonstrates enhanced ability to handle complex, multi-step workflows. On the GDPval-AA benchmark, K3.1 Pro reached an Elo rating of 1,846, surpassing its predecessor by 312 points. In coding environments, the model scored 65.7 percent on Terminal-Bench 4.0, a massive leap from the 13 percent recorded by K3. It also reduced hallucination rates in Moonshot’s internal stress tests from 23.4 percent to 8.3 percent. While GPT-6 Astra still holds an edge with a 5.5 percent rate, the reduction in K3.1 Pro represents a substantial reliability upgrade for enterprise use cases.

Beyond standard text and code generation, the model has expanded its capabilities into 3D modeling. It now works directly with game engines like Unity, Unreal, and Godot, and can write scripts for Blender and CAD software without requiring additional connectors. On the BenchCAD benchmark, it achieved 92.7 percent with tools and 72.8 percent without, performing nearly identically to top-tier competitors. For users, this means the ability to preview, edit, and export 3D models directly through the Kimi.com interface.

Despite these performance gains, Moonshot has made the model more accessible by lowering API prices by 33 percent. Input tokens now cost $2 per million, down from $3, while output tokens are $10, down from $15. The model retains the same 2.8 trillion total parameters and 104 billion active parameters as its predecessor, maintaining a one-million-token context window. This combination of high performance and lower cost places it at the top of the Pareto frontier for models of its size.

Looking ahead, Moonshot is preparing for the open-weights release scheduled for early November. The company has taken deliberate steps to limit offensive cybersecurity capabilities by stripping cyber-related data from the training process. While the uncleaned version solved 90.5 percent of tasks in their internal vulnerability benchmark, the cleaned K3.1 Pro solves 81.0 percent. This strategic balance allows the model to serve defensive purposes, such as auditing legacy systems, without pushing the absolute frontier of offensive cyber threats. Users can access the model now through Kimi.ai, Kimi Work, and the Kimi API.

The arrival of K3.1 Pro signals that the distinction between open and proprietary AI is blurring. By delivering top-tier reasoning at a reduced price point, Moonshot is forcing competitors to justify their premium pricing with even stronger performance. As the open-weights version becomes available, expect to see a wider adoption of these capabilities in developer workflows, further accelerating the pace of innovation in the AI sector.