AI Race: Kimi K3.1 Pro Narrows Gap With Top Proprietary Models
Moonshot AI’s latest release scores 55 points on key benchmarks, challenging the dominance of leading competitors in the artificial intelligence landscape.
By Ana Castro ·
A score of 55 on the Artificial Analysis Intelligence Index signals a significant shift in the global artificial intelligence hierarchy. Moonshot AI recently launched Kimi K3.1 Pro, a post-trained iteration of its previous model that has surged to fourth place. This 11-point increase over its predecessor places it just three points behind the industry leaders, marking a notable reduction in the performance gap between Chinese and American AI systems.
The primary driver of this advancement is the model’s enhanced capability in agentic tasks. On the GDPval-AA benchmark, Kimi K3.1 Pro achieved an Elo rating of 1,846, a jump of 312 points from the earlier version. It now trails only Claude Opus 5.5 and Muse Wonder Full in this specific metric. Furthermore, the model demonstrated substantial improvements in coding and terminal operations, recording a 65.7 percent success rate on Terminal-Bench 4.0. This figure surpasses GPT-6 Astra and approaches the 66.4 percent recorded by the leading competitor.
Beyond raw performance, Moonshot has optimized the economic structure of the model. While maintaining the same 2.8 trillion total parameters and 104 billion active parameters as its predecessor, the company reduced API pricing by 33 percent. Input costs now stand at $2 per million tokens, while output costs are $10. The company also noted that hallucination rates dropped significantly on internal stress tests, falling from 23.4 percent to 8.3 percent, although GPT-6 Astra retains a slight edge with a 5.5 percent error rate in the same evaluation.
The release also highlights progress in specialized fields such as 3D modeling and cybersecurity. Kimi K3.1 Pro can interface directly with game engines like Unity and Unreal Engine, scoring 92.7 percent on the BenchCAD benchmark when using external tools. Moonshot has also introduced a feature on its platform that allows users to preview and export 3D models. In the realm of cybersecurity, the company stripped cyber-related data from the training process for the open-weights version to balance offensive and defensive capabilities, resulting in an 81.0 percent solve rate on internal vulnerability reproduction tests.
Contextually, this update positions Kimi K3.1 Pro as a potential leader in the open-weights category. If the model weights are released by November 4, 2026, as planned, it would become the highest-scoring open model available. This move would narrow the disparity between open and proprietary systems from 12 points to just 3. The company has also introduced a Highspeed variant that runs two to three times faster, expanding its utility for developers needing rapid inference. With the model now available through various Kimi platforms, the competitive pressure on established players is expected to intensify in the coming months.
