Moonshot's New Model Challenges AI Giants
Kimi K3.1 Pro has surged to fourth place in intelligence rankings, significantly narrowing the performance gap with leading proprietary systems.
By Fatima Zayed ·
Moonshot AI has just released Kimi K3.1 Pro, a significant upgrade to its previous model that has immediately reshaped the competitive landscape in artificial intelligence. The new system has climbed to the fourth spot on the Artificial Analysis Intelligence Index, scoring 55 points. This places it just three points behind the current top proprietary models, marking a substantial leap forward for a company based in China.
The most immediate impact of this release is its potential to redefine the open-weights market. If Moonshot follows through on its plan to release the model's weights by November 4, 2026, Kimi K3.1 Pro would become the highest-scoring open-weights model available. This development would shrink the performance gap between open and proprietary models from 12 points to just three, assuming no other major releases occur in the interim.
One of the primary drivers of this performance jump is the model's enhanced capability in agentic tasks, which involve autonomous problem-solving and execution. On the GDPval-AA benchmark, the new model achieves an Elo score of 1,846, a 312-point increase over its predecessor. It now trails only Claude Opus 5.5 and Muse Wonder Full in this specific category. Furthermore, the model shows dramatic improvement in terminal-based coding tasks, where it scores 65.7 percent on Terminal-Bench 4.0, a massive rise from the previous version's 13 percent score.
Beyond raw coding ability, the model demonstrates better reliability and specialized knowledge. Moonshot reports that hallucination rates have dropped sharply on internal stress tests, falling from 23.4 percent to 8.3 percent. While GPT-6 Astra still holds a slight edge with a 5.5 percent rate, the reduction indicates a more stable and accurate system. The model also shows gains in financial analysis, topping all tested models on office task benchmarks and improving scores in financial research and modeling.
A notable new feature is the model's integration with 3D design and game development workflows. It can interact directly with engines like Unity, Unreal Engine, and Godot to handle both minor fixes and full project builds. On the BenchCAD benchmark, it achieves a 92.7 percent score when using tools, closely matching top competitors. Moonshot has also introduced a "3D Make" feature on its platform, allowing users to preview, edit, and export 3D models generated by the AI.
Despite the performance boost, the model maintains the same architectural size as its predecessor, with 2.8 trillion total parameters and 104 billion active ones. However, the cost of using the model has decreased. Input costs are now $2 per million tokens, down from $3, while output costs have fallen to $10 from $15. This price reduction represents a 33 percent drop, making the advanced capabilities more accessible to developers and businesses.
Moonshot has also addressed cybersecurity concerns, a critical factor for open-weights models. The company removed cyber-related data from the training process to prevent the model from becoming too potent in offensive security tasks. On internal tests that reproduce recent software vulnerabilities, the cleaned model solves 81.0 percent of tasks, down from 90.5 percent in the uncleaned version. This balance aims to keep the model useful for defensive auditing without pushing the absolute frontier of offensive capabilities.
The release includes a "Highspeed" variant that runs two to three times faster, catering to users who need quick responses. Both versions are available immediately through various Moonshot platforms and APIs. As the AI industry continues to evolve, this release highlights how rapidly open models are closing the gap with proprietary leaders, potentially shifting the balance of power in the technology sector.
Source: The Decoder
