Moonshot AI Launches Kimi K3.1 Pro to Challenge Top US AI Models

Chinese AI developer Moonshot AI has released Kimi K3.1 Pro, a new reasoning model that immediately moved into fourth place on the Artificial Analysis Intelligence Index. The update closes the distance between open and proprietary systems, scoring just three points behind the current leaders. This places the model in close competition with major industry players and signals a shift in the global AI landscape.

The new system scores 55 points on the index, marking an 11 point increase over its predecessor, Kimi K3. Currently, only Anthropic’s Claude Opus 5.5, Claude Sonnet 5.5, and Muse Wonder Full hold higher scores. Moonshot plans to release the model weights by November 4, 2026. If released as scheduled, this would make it the top performing open weights model available and reduce the gap with proprietary leaders from 12 points to just three.

Performance gains are most evident in agentic tasks and coding. On the GDPval-AA benchmark, the model reached an Elo of 1,846, surpassing the previous version by 312 points. It trails only Claude Opus 5.5 and Muse Wonder Full in this specific metric. In coding tests, the model scored 65.7 percent on Terminal-Bench 4.0. This result beats GPT-6 Astra and lands just below Claude Opus 5.5. The previous K3 model managed only 13 percent on that same test.

Moonshot also reports a significant reduction in hallucinations. In internal testing, claim-level errors dropped from 23.4 percent to 8.3 percent. The company notes that this benchmark serves as a stress test rather than a measure of general accuracy. GPT-6 Astra still leads this internal metric with a 5.5 percent error rate. The new model also shows strong gains in finance, topping tested systems on OfficeQA Pro v2 and improving scores in financial modeling and research.

The architecture remains consistent with the previous version, featuring 2.8 trillion total parameters and a one million token context window. API prices have dropped by 33 percent, with input tokens costing $2 per million. Despite generating 14 percent more output tokens, the cost to run each index task remains nearly identical to the previous model. This efficiency makes the system more accessible for developers and businesses.

The company highlighted improvements in 3D modeling and cybersecurity. Kimi K3.1 Pro can work with game engines like Unity and Unreal Engine without special connectors. It also performs well on cybersecurity benchmarks, though the company removed specific cyber data from the open weights release to balance offensive and defensive capabilities. This approach aims to allow open models to support defensive auditing while gradually raising their overall security capabilities.

Users can access both the standard and high-speed versions of the model through various platforms, including the Kimi API and web interface. The high-speed variant is designed to run two to three times faster. This release positions Moonshot AI as a key competitor in the high-end AI market, offering a powerful alternative to established proprietary systems.