Kimi K3.1 Pro narrows the gap with top AI models

55 points. That is the score Moonshot AI’s new Kimi K3.1 Pro model has just achieved on the Artificial Analysis Intelligence Index. This result places the Chinese-developed system in fourth position, a narrow margin behind the current leaders. The primary competitor, Anthropic’s Claude Opus 5.5, sits at the top with 58 points. Claude Sonnet 5.5 and Muse Wonder Full both score 56, keeping them ahead of the new entrant.

The numbers suggest a significant shift in the competitive landscape. K3.1 Pro represents an 11-point improvement over its predecessor, K3. If Moonshot follows through on its plan to release the model weights by November 4, 2026, it will become the highest-scoring open-weights model available. This move would shrink the performance gap between open and proprietary systems from 12 points to just three, assuming no other major releases occur in the interim.

The company has also introduced a variant named K3.1 Pro Highspeed. According to Moonshot, this version runs two to three times faster than the standard model. Both options are currently accessible through Kimi.ai, Kimi Work, Kimi Code, and the Kimi API, giving developers and users immediate access to the upgraded capabilities.

Performance gains are most evident in agentic tasks and coding. On the GDPval-AA benchmark, the new model reaches an Elo rating of 1,846, a substantial 312-point increase from K3. It trails only Claude Opus 5.5 and Muse Wonder Full in this specific metric. In coding tests, K3.1 Pro reports a 65.7 percent success rate on Terminal-Bench 4.0. This figure surpasses GPT-6 Astra and closely follows Claude Opus 5.5. Notably, the older K3 model managed only 13 percent on the same test.

The model also handles 3D modeling and financial tasks with greater precision. It can work directly with game engines like Unity and Unreal Engine, scoring 92.7 percent on the BenchCAD benchmark when using tools. Hallucination rates have dropped significantly as well. On an internal stress test, claim-level hallucinations fell from 23.4 percent with K3 to 8.3 percent with K3.1 Pro, though GPT-6 Astra still holds an edge at 5.5 percent.

Pricing has become more competitive alongside the performance boost. Moonshot has cut API costs by 33 percent. Input tokens now cost $2 per million, down from $3, while output tokens are $10, reduced from $15. Despite the lower price, the cost per Intelligence Index task remains nearly identical to the previous version at $1.98, due to the model generating 14 percent more output tokens.

For the upcoming open-weights release, Moonshot has taken deliberate steps regarding cybersecurity capabilities. The company removed cyber-related data from both training phases to manage offensive potential. On an internal benchmark for reproducing software vulnerabilities, the uncleaned version solved 90.5 percent of tasks, while the cleaned K3.1 Pro solved 81.0 percent. This approach aims to balance offensive capability with defensive utility for auditing legacy systems.