Moonshot AI Launches Kimi K3.1 Pro, Narrowing Gap With Top US Models

Moonshot AI has released Kimi K3.1 Pro, a new reasoning model that immediately secured the fourth-highest ranking on the Artificial Analysis Intelligence Index. The release marks a significant milestone for Chinese AI development, as the new system narrows the performance gap with leading proprietary models from United States-based companies to a mere three points. This positioning places Kimi K3.1 Pro directly behind Claude Opus 5.5, Claude Sonnet 5.5, and Muse Wonder Full, which currently hold the top three spots with scores of 58, 56, and 56, respectively.

The new model represents a substantial upgrade over its predecessor, Kimi K3. By applying significantly more post-training compute, Moonshot AI achieved an 11-point increase in its intelligence index score, jumping from 44 to 55. If the company follows through on its plan to release the model's weights by November 4, 2026, Kimi K3.1 Pro is set to become the highest-scoring open-weights model available. Such a move would reduce the performance disparity between open and proprietary models from 12 points to just three, assuming no other major releases occur in the interim.

Alongside the standard version, Moonshot is shipping a variant called K3.1 Pro Highspeed. According to the company, this version operates two to three times faster than its counterpart. Both models are currently accessible through Kimi.ai, Kimi Work, Kimi Code, and the Kimi API, allowing developers and users to integrate them into their workflows immediately.

The most notable improvements in Kimi K3.1 Pro appear in agentic tasks, which involve the model performing multi-step actions to achieve specific goals. On the GDPval-AA benchmark, the new model reached 1,846 Elo, a 312-point leap over the previous version. This score places it just behind Claude Opus 5.5 and Muse Wonder Full. In terminal-based coding tests, specifically Terminal-Bench 4.0, the model achieved a 65.7 percent success rate. This figure surpasses GPT-6 Astra and comes within a narrow margin of Claude Opus 5.5, while the older Kimi K3 managed only 13 percent on the same evaluation.

Reliability has also seen a marked improvement. Moonshot reports that hallucination rates dropped sharply on its internal Kimi Hallu Bench Hard. Claim-level hallucinations fell from 23.4 percent in the previous model to 8.3 percent. While the company notes this is a stress test rather than a measure of general error rates, the reduction is significant. GPT-6 Astra still holds an advantage on this specific benchmark with a 5.5 percent rate. In financial applications, the model showed gains in both modeling and research, topping all tested systems on general office tasks.

The model also demonstrates strong capabilities in 3D modeling and game development. It can interact directly with engines such as Unity, Unreal Engine, and Godot, handling everything from minor fixes to full project builds. On the BenchCAD benchmark, Kimi K3.1 Pro scored 92.7 percent when using tools, approaching the performance of top US models. Without external tools, it achieved 72.8 percent, nearly matching Claude Opus 5.5. The company has highlighted its strength in parametric modeling of mechanical parts and architectural structures, and has launched a "3D Make" feature on its website for users to preview and export models.

Despite the performance gains, the model retains the same architectural dimensions as Kimi K3, featuring 2.8 trillion total parameters and 104 billion active parameters. It supports a context window of one million tokens and processes both text and images. Artificial Analysis notes that the model sits at the top of the Pareto frontier, meaning no other model with a comparable parameter count achieves a higher intelligence index score.

Cost efficiency has improved alongside performance. API prices have dropped by 33 percent, with input costs now at $2 per million tokens and output at $10. Cache hits cost $0.20. The total cost per intelligence index task remains nearly identical to the previous model at $1.98, as the new system generates 14 percent more output tokens during tasks.

Moonshot also addressed cybersecurity considerations in its technical documentation. The company observed that vulnerability research capabilities grew as post-training scaled, due to cyber-related content in the training corpus. For the open-weights release, they removed this data to limit offensive capabilities. On internal tests reproducing recent vulnerabilities, the uncleaned version solved 90.5 percent of tasks, while the cleaned K3.1 Pro solved 81.0 percent. This is higher than the previous model's 33.3 percent but below DeepSeek-V4.1-Flash at 85.7 percent. The company aims to balance offensive potential with defensive utility, such as auditing legacy systems.