Moonshot AI Unveils Kimi K3.1 Pro, Narrowing Gap with Leading Proprietary Models

Moonshot AI has officially launched Kimi K3.1 Pro, a post-trained reasoning model that has rapidly ascended the global rankings. The release marks a significant shift in the competitive landscape, positioning the Chinese firm’s latest creation just three points behind the top proprietary systems in current evaluations.

On the Artificial Analysis Intelligence Index, K3.1 Pro scores 55 points. This represents an 11-point increase over its predecessor, Kimi K3, securing the model the fourth-highest ranking. Only Claude Opus 5.5, which holds 58 points, and two other models tied at 56 points rank higher. If the company adheres to its schedule, this performance will make it the highest-scoring open-weights model available.

The model maintains the same architectural scale as its predecessor, featuring 2.8 trillion total parameters with 104 billion active. It supports a context window of one million tokens and processes both text and image inputs. According to Artificial Analysis, K3.1 Pro sits at the top of the Pareto frontier, meaning no other model of comparable size achieves a higher score on the intelligence index.

Moonshot AI intends to release the full model weights by November 4, 2026. If executed as planned, this move would reduce the performance gap between the best open and proprietary models from 12 points to just three. The company also introduced a variant named K3.1 Pro Highspeed, which is designed to run two to three times faster than the standard version. Both models are currently accessible through Kimi.ai, Kimi Work, Kimi Code, and the Kimi API.

Performance gains are most pronounced in agentic tasks and coding. On the GDPval-AA benchmark, the model reaches an Elo rating of 1,846, a substantial 312-point improvement over K3. In terminal-based coding tests, K3.1 Pro achieves a 65.7 percent success rate on Terminal-Bench 4.0, surpassing GPT-6 Astra and closely trailing Claude Opus 5.5. The model also demonstrates improved reliability in financial modeling and general office tasks, topping all tested models on the OfficeQA Pro v2 benchmark.

In the realm of 3D modeling, K3.1 Pro can interact directly with game engines such as Unity, Unreal Engine, and Godot. It scores 92.7 percent on the BenchCAD benchmark when using tools, a result close to the leading proprietary models. Moonshot notes that the system is particularly strong in parametric modeling of mechanical parts and architectural structures, capable of writing and executing scripts for Blender and CAD software without the need for additional connectors.

The launch coincides with a 33 percent reduction in API pricing. Input costs have dropped to $2 per million tokens, while output costs are now $10 per million tokens. Cache hits are priced at $0.20 per million tokens. Despite the price cuts, the cost per intelligence index task remains nearly identical to the previous version at $1.98, a result of the model generating 14 percent more output tokens.

Addressing cybersecurity concerns, Moonshot AI has removed cyber-related data from the training phases of the open-weights version. In internal testing, the uncleaned model solved 90.5 percent of vulnerability reproduction tasks, whereas the cleaned K3.1 Pro solves 81.0 percent. The company aims to gradually expand the defensive capabilities of open models while ensuring they do not push the absolute frontier of offensive cyber operations.

The release signals an intensifying race in the artificial intelligence sector, with Chinese developers increasingly matching the performance of leading US-based competitors. By combining competitive benchmark scores with lower costs and a planned open-weights release, Moonshot AI is positioning itself as a key player in the next phase of model development.