New Chinese AI model cuts prices and narrows gap with top rivals

For developers and businesses relying on artificial intelligence tools, the cost of running complex tasks has just dropped significantly. Moonshot AI, a major Chinese tech firm, has launched Kimi K3.1 Pro, a new reasoning model that undercuts the prices of its predecessor while delivering sharper performance in specialized fields like software development and 3D design. The update also positions the model as a strong competitor to the leading systems from US-based companies.

The new model has jumped to fourth place on the Artificial Analysis Intelligence Index with a score of 55. It trails only three proprietary models: Claude Opus 5.5 at 58 points, and Claude Sonnet 5.5 and Muse Wonder Full, both at 56. If Moonshot releases the model's weights by November 4, 2026, as planned, K3.1 Pro 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.

The most notable improvements appear in agentic tasks, where the model can execute multi-step workflows. On the GDPval-AA benchmark, K3.1 Pro reached 1,846 Elo, a substantial increase over its predecessor. In coding tests, it recorded a score of 65.7 percent on Terminal-Bench 4.0, surpassing GPT-6 Astra and coming very close to the leader, Claude Opus 5.5. The model also handles 3D modeling with greater precision, capable of working directly with game engines like Unity and Unreal Engine. It can write and run scripts for design software without needing additional connectors, a feature Moonshot highlights as a key advantage for creative professionals.

While performance has risen, the financial burden for users has decreased. API prices have fallen by 33 percent, with input costs now at $2 per million tokens and output at $10. This is a significant reduction from the previous rates of $3 and $15, respectively. The model retains its large scale, featuring 2.8 trillion total parameters and a one-million token context window. Despite the price cut, the cost per intelligence index task remains nearly identical to the previous version because the new model generates slightly more output tokens to achieve its results.

Moonshot has also taken specific steps to address cybersecurity concerns. For the open-weights release, the company removed cyber-related data from the training process to prevent the model from becoming overly specialized in offensive tasks. On internal tests that reproduce known software vulnerabilities, the cleaned version solves 81.0 percent of tasks, compared to 90.5 percent for the uncleaned version. The company argues this balance allows the model to be useful for defensive auditing without pushing the absolute limits of offensive capability.

A faster variant, K3.1 Pro Highspeed, is also available, running two to three times quicker than the standard model. Both versions are accessible through Kimi's various platforms and API services. The release marks a significant shift in the competitive landscape, as Chinese models continue to close the performance gap with top US rivals while offering more competitive pricing for global users.

The next major milestone for this model is the release of its weights in early November. If this timeline holds, developers will have access to a top-tier open-weights system that performs nearly on par with the most advanced proprietary models. This development could accelerate the adoption of open models in enterprise environments, where cost and customization are often critical factors.