Moonshot's New AI Model Cuts Prices and Narrows the Gap With Top Rivals

Moonshot AI has just released Kimi K3.1 Pro, a new reasoning model that has immediately reshaped the competitive landscape. The system currently holds the fourth spot on the Artificial Analysis Intelligence Index. It sits just three points behind the leading proprietary models. This places it in close contention with Anthropic's Claude Opus 5.5.

The jump to this rank represents an 11-point increase over its predecessor, Kimi K3. According to the tracking index, only three models currently score higher. These are Claude Opus 5.5, Claude Sonnet 5.5, and Muse Wonder Full. If Moonshot releases the model weights as planned by November 4, 2026, K3.1 Pro would become the highest-scoring open-weights model. That move would shrink the performance gap between open and proprietary systems from 12 points to just three, assuming no other changes occur.

The model shows its strongest improvements in agentic tasks. On the GDPval-AA benchmark, it reaches an Elo score of 1,846. This is a 312-point rise over the previous version. It now trails only Claude Opus 5.5 and Muse Wonder Full in that category. On Terminal-Bench 4.0, the model achieves a 65.7 percent success rate. This beats GPT-6 Astra, which scored 58.2 percent. It also barely falls short of Claude Opus 5.5, which recorded 66.4 percent. The earlier Kimi K3 managed only 13 percent on this same test.

Reliability has also improved significantly. Internal testing shows that claim-level hallucinations dropped from 23.4 percent to 8.3 percent. Moonshot notes this is a stress test rather than a measure of general error rates. GPT-6 Astra still holds an edge at 5.5 percent on the same benchmark. In finance, the new model gained 7.2 points in modeling and 11.6 points in research. For general office work, it tops all tested models on the OfficeQA Pro v2 benchmark.

Coding and 3D modeling capabilities have expanded as well. The model works directly with game engines like Unity, Unreal Engine, and Godot. It handles both minor fixes and full project builds. On the BenchCAD benchmark, it scores 92.7 percent with tools. This is close to the scores of Claude Opus 5.5 and GPT-6 Astra. Without tools, it hits 72.8 percent, nearly matching the top rival. The company says it excels at parametric modeling for mechanical parts and architecture. It can write and run scripts for Blender and CAD software without needing specific connectors.

Despite the performance gains, the model size remains the same at 2.8 trillion total parameters. It uses 104 billion active parameters and supports a one-million-token context window. It handles both text and images. According to Artificial Analysis, it sits at the top of the Pareto frontier. No model with a comparable size scores higher on the Intelligence Index. A faster variant, K3.1 Pro Highspeed, is also available. It runs two to three times faster than the standard version.

Pricing has dropped by 33 percent. Input costs are now $2 per million tokens, down from $3. Output costs fall to $10 from $15. Cache hits cost $0.20, reduced from $0.30. The cost per Intelligence Index task is $1.98, nearly identical to the previous version. This is because the new model generates 14 percent more output tokens.

Moonshot also addressed cybersecurity concerns in its technical documentation. The company removed cyber-related data from training phases for the open-weights release. This reduces the model's offensive capabilities slightly. On an internal benchmark, the uncleaned version solved 90.5 percent of vulnerability tasks. After cleaning, K3.1 Pro solves 81.0 percent. DeepSeek-V4.1-Flash scores 85.7 percent on the same test. The company aims to balance offensive potential with defensive utility for auditing legacy systems.