Moonshot AI Launches Kimi K3.1 Pro Open-Weight Model
Moonshot AI released Kimi K3.1 Pro, a 3-trillion-parameter model with enhanced coding and financial research capabilities and restricted cyber-offensive features.
Beijing-based startup Moonshot AI launched Kimi K3.1 Pro on October 24, 2026, a 3-trillion-parameter open-weight model featuring a 1-million-token context window. The model is designed to improve performance in coding, 3D creation, and financial research. On the Terminal-Bench 4.0 coding test, Kimi K3.1 Pro scored 65.7%, surpassing OpenAI's GPT-6 Astra and nearly matching Anthropic's Claude Opus 5.5.
The company also introduced a Highspeed variant that operates two to three times faster than the standard version. Both are currently available via Kimi apps and API, with full model weights scheduled for release by November 4 to allow for developer customization. The launch comes as the company targets $2 billion in annualized income by the end of 2026 and considers a potential Hong Kong listing in 2027.
To address cybersecurity concerns, Moonshot AI stripped cyber-related data from the model's training sets and reinforcement-learning environments. Internal benchmarks showed this reduced the model's ability to exploit software vulnerabilities from 90.5% to 81.0%. This move aligns with proposed export controls from the Government of China to ban AI-driven malware generation. The release follows previous accusations from the White House that the company used banned Nvidia chips and copied an Anthropic model.
