AI Pricing Drops as New Chinese Model Rivals Industry Leaders

Developers and businesses relying on artificial intelligence APIs will see a noticeable drop in costs this week. Moonshot AI has launched Kimi K3.1 Pro, a new reasoning model that reduces token prices by 33 percent. This financial shift occurs alongside a significant performance upgrade, positioning the new model as a serious competitor to established industry leaders.

The new model scores 55 points on the Artificial Analysis Intelligence Index, marking an 11-point increase over its predecessor, Kimi K3. This performance places it in fourth position globally. Only three models currently rank higher: Claude Opus 5.5 with 58 points, and both Claude Sonnet 5.5 and Muse Wonder Full with 56 points. The narrow three-point gap between K3.1 Pro and the leaders signals a tightening race at the top of the AI landscape.

Moonshot AI is the developer behind this technology. The company is releasing two versions of the model. The standard K3.1 Pro is available immediately, as is a variant called K3.1 Pro Highspeed, which the company says operates two to three times faster. Both models are accessible through Kimi.ai, Kimi Work, Kimi Code, and the Kimi API. Users can interact with the system using both text and image inputs, supported by a one-million-token context window.

The stakes for the AI industry are high, particularly regarding the distinction between open and proprietary models. Moonshot plans to release the model weights by November 4, 2026. If this timeline holds, K3.1 Pro would become the highest-scoring open-weights model available. This release would shrink the performance gap between open and proprietary systems from 12 points to just three, assuming no other models change their status in the interim.

Performance gains are most evident in agentic tasks and coding. On the GDPval-AA benchmark, the model achieves 1,846 Elo, a 312-point jump from the previous version. It trails only Claude Opus 5.5 and Muse Wonder Full on this specific test. In coding, Moonshot reports a 65.7 percent score on Terminal-Bench 4.0, surpassing GPT-6 Astra. The model also demonstrates improved reliability, with internal hallucination rates dropping from 23.4 percent to 8.3 percent.

Despite the performance boosts, the model retains its 2.8 trillion total parameters, with 104 billion active. Input costs now stand at $2 per million tokens, down from $3, while output costs fall to $10 from $15. To manage cybersecurity risks, Moonshot removed specific cyber-related data from the training process before the open-weights release. This adjustment reduced its ability to solve recent vulnerability tasks from 90.5 percent to 81.0 percent, a deliberate trade-off to balance offensive and defensive capabilities.

The model also shows strong results in 3D modeling, scoring 92.7 percent on BenchCAD with tools. This is nearly on par with top competitors. Moonshot notes that the system can write and run scripts for Blender and CAD software without additional connectors. A new feature called '3D Make' allows users to preview, edit, and export 3D models directly on the platform. This release represents a significant step for Chinese AI firms in closing the performance gap with American counterparts.