Moonshot AI's Kimi K3.1 Pro Narrows Gap With Top Rivals

55. That is the score Kimi K3.1 Pro achieved on the Artificial Analysis Intelligence Index, a result that places the new Chinese model in fourth place globally. It trails only Claude Opus 5.5, Claude Sonnet 5.5, and Muse Wonder Full. This performance represents a significant leap for Moonshot AI, marking an 11-point increase over its predecessor, Kimi K3.

The release follows a steady escalation in competition among frontier AI developers. Moonshot developed K3.1 Pro as a post-trained iteration of Kimi K3, utilizing significantly more post-training compute to refine the model's reasoning capabilities. The company has made both the standard K3.1 Pro and a variant called K3.1 Pro Highspeed available immediately. The Highspeed version is designed to run two to three times faster than the standard release, according to Moonshot.

The most dramatic improvements are evident in agentic tasks, where the model demonstrates enhanced autonomy and precision. On the GDPval-AA benchmark, K3.1 Pro recorded 1,846 Elo, a jump of 312 points over K3. This places it slightly behind Claude Opus 5.5 and Muse Wonder Full. In coding evaluations, the model scored 65.7 percent on Terminal-Bench 4.0, surpassing GPT-6 Astra and closely trailing Claude Opus 5.5. Notably, the previous K3 version managed only 13 percent on this same test.

Moonshot also reports a sharp decline in hallucination rates. On its internal Kimi Hallu Bench Hard, claim-level hallucinations dropped from 23.4 percent in K3 to 8.3 percent in K3.1 Pro. While GPT-6 Astra maintains a lower rate of 5.5 percent, Moonshot describes its internal test as a stress test rather than a general error metric. The model also showed gains in financial modeling and research, and it tops all tested models on the OfficeQA Pro v2 benchmark for general office tasks.

In specialized fields like 3D modeling, K3.1 Pro integrates directly with game engines such as Unity, Unreal Engine, and Godot. It achieved 92.7 percent on the BenchCAD benchmark when using tools, a figure close to the leading proprietary models. The company launched a feature called "3D Make" on Kimi.com, allowing users to preview, edit, and export 3D models. The model is particularly strong at parametric modeling of mechanical parts and architectural designs.

Despite the performance boost, the model retains the same architecture as K3, with 2.8 trillion total parameters and 104 billion active ones. It supports a one-million-token context window and processes both text and images. However, API prices have dropped by 33 percent. Input costs are now $2 per million tokens, down from $3, while output costs stand at $10, reduced from $15. Cache hits are priced at $0.20 per million tokens. This efficiency places the model at the top of the Pareto frontier, meaning no comparable model scores higher on the Intelligence Index.

Looking ahead, Moonshot plans to release the model weights by November 4, 2026. If executed as planned, K3.1 Pro would become the highest-scoring open-weights model available. This move would shrink the gap between open and proprietary models from 12 points to just three. The company also detailed its approach to cybersecurity, noting that it stripped cyber-related data from training phases for the open-weights release to balance offensive and defensive capabilities.