Kimi K3.1 Pro Slashes AI Costs by a Third While Closing the Gap with Top Proprietary Models

A 33 percent reduction in token pricing has accompanied the launch of Kimi K3.1 Pro, a new reasoning model from Chinese AI developer Moonshot AI that is rapidly narrowing the distance between open-source and proprietary artificial intelligence systems. The release marks a significant shift in the competitive landscape, as the model achieves a score of 55 on the Artificial Analysis Intelligence Index, placing it in fourth position globally.

The new model represents a substantial leap from its predecessor, Kimi K3, with an 11-point increase in its overall intelligence score. It now trails only three models: Claude Opus 5.5, which holds the top spot with 58 points, and both Claude Sonnet 5.5 and Muse Wonder Full, which sit at 56 points each. If Moonshot adheres to its plan to release the model weights by November 4, 2026, K3.1 Pro will become the highest-scoring open-weights model available. This move would shrink the performance gap between the best open models and the top proprietary systems from 12 points to just three.

The performance gains are most evident in agentic tasks, where the model has made dramatic strides. On the GDPval-AA benchmark, K3.1 Pro reaches an Elo score of 1,846, a 312-point improvement over the previous version. This places it just behind Claude Opus 5.5 and Muse Wonder Full. In coding and terminal-based tasks, the model reports a 65.7 percent success rate on Terminal-Bench 4.0, a stark contrast to the 13 percent achieved by K3. It also outperforms GPT-6 Astra, which scored 58.2 percent on the same test, while remaining slightly behind Claude Opus 5.5 at 66.4 percent.

Moonshot AI has also highlighted significant improvements in the model's reliability. Internal testing on the Kimi Hallu Bench Hard shows that claim-level hallucinations dropped from 23.4 percent in the previous version to 8.3 percent. While the company notes this is a stress test rather than a general error metric, GPT-6 Astra maintains a lead with a 5.5 percent rate on the same benchmark. In financial applications, the model improved by 7.2 points in modeling and 11.6 points in research, and it currently tops all tested models on the OfficeQA Pro v2 benchmark for general office tasks.

The model’s capabilities extend into specialized fields such as 3D modeling and game development. It can interact directly with engines like Unity, Unreal, and Godot for both quick fixes and full project builds. On the BenchCAD benchmark, K3.1 Pro achieves 92.7 percent with tools, closely following Claude Opus 5.5 and GPT-6 Astra. Without tools, it scores 72.8 percent, nearly matching the top proprietary model. Moonshot states the model is particularly strong in parametric modeling for mechanical parts and architecture, and it can write scripts for Blender and CAD software without requiring additional connectors. A new feature called "3D Make" on Kimi.com allows users to preview and export these models.

Despite the performance boost, the model retains the same architectural dimensions as its predecessor, with 2.8 trillion total parameters and 104 billion active ones. It supports a one-million-token context window and handles both text and images. According to Artificial Analysis, it sits at the top of the Pareto frontier, meaning no other model with a comparable parameter count scores higher. The price drop to $2 per million input tokens and $10 per million output tokens makes it a cost-effective option, with each Intelligence Index task costing approximately $1.98.

Moonshot AI has also addressed cybersecurity concerns in its technical release. The company observed that vulnerability research capabilities grew as post-training scaled, so it stripped cyber-related data from the training phases for the open-weights version. On an internal benchmark reproducing recent vulnerabilities in software like the Linux kernel, the uncleaned model solved 90.5 percent of tasks, while the cleaned version solved 81.0 percent. This compares to 85.7 percent for DeepSeek-V4.1-Flash and 33.3 percent for the previous Kimi K3. The company aims to gradually raise the capabilities of open models for defensive purposes while ensuring they do not push the absolute frontier of offensive cyber capabilities.

Both the standard K3.1 Pro and a faster variant called K3.1 Pro Highspeed are now available through Kimi.ai, Kimi Work, Kimi Code, and the Kimi API. The Highspeed version runs two to three times faster than the standard release. The upcoming release of the model weights in November is expected to intensify competition in the AI sector, as developers gain access to a high-performing, cost-effective open-weights option that challenges the dominance of established proprietary systems.