Moonshot AI’s Kimi K3.1 Pro Challenges Industry Leaders With Lower Costs
A new Chinese AI model has surged up global rankings while reducing API prices by a third, intensifying the competition between open and proprietary systems.
By Priya Sharma ·
Kimi K3.1 Pro, the latest reasoning model from Chinese developer Moonshot AI, has arrived as a serious challenger to the top-tier proprietary models currently dominating the market. By securing the fourth spot on the Artificial Analysis Intelligence Index, the new system has narrowed the performance gap between open-weights and closed models to just three points, a significant shift from the twelve-point difference seen previously.
The model achieves a score of 55 on the index, an eleven-point increase over its predecessor, Kimi K3. 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 releases the model weights as planned by November 4, 2026, K3.1 Pro will become the highest-scoring open-weights model available, assuming no other releases occur in the interim.
Performance gains are most visible in agentic tasks and coding. On the GDPval-AA benchmark, the new model scores 1,846 Elo, a 312-point jump from K3, placing it just behind Claude Opus 5.5. In terminal-based coding tests, it reported a 65.7 percent success rate, far exceeding the 13 percent recorded by the previous version. Additionally, hallucination rates on Moonshot’s internal stress test dropped from 23.4 percent to 8.3 percent, though GPT-6 Astra remains slightly lower at 5.5 percent.
The model also handles 3D modeling with notable efficiency. It can work directly with game engines like Unity and Unreal, scoring 92.7 percent on the BenchCAD benchmark when using tools. Moonshot notes that the system is particularly strong at parametric modeling for mechanical and architectural parts, capable of writing and running scripts for Blender without additional connectors. A new feature called "3D Make" now allows users to preview and export models directly on the Kimi platform.
Despite the performance leap, the model maintains the same 2.8 trillion total parameters as its predecessor, with 104 billion active. However, API prices have dropped by 33 percent. Input costs are now $2 per million tokens, down from $3, while output costs have fallen to $10 from $15. This places the model at the top of the Pareto frontier, meaning no other model with a similar parameter count scores higher on the intelligence index.
Moonshot has also implemented deliberate limits on offensive cybersecurity capabilities for the open-weights release. By removing specific cyber-related data from the training process, the model’s ability to solve hard vulnerability tasks dropped from 90.5 percent to 81.0 percent. While this is still higher than the previous K3 model’s 33.3 percent, the company aims to balance defensive utility with controlled offensive potential.
A high-speed variant, K3.1 Pro Highspeed, is also available and runs two to three times faster than the standard version. Both models are accessible now through Kimi.ai, Kimi Work, Kimi Code, and the API, offering a more affordable entry point into high-performance agentic AI.
Source: The Decoder
