According to reports, the AI programming tool Cursor has officially released its new-generation intelligent coding model, Composer. Utilizing reinforcement learning and a Mixture of Experts (MoE) architecture, the model achieves a code generation speed of 250 tokens per second—approximately four times faster than similar models—with most tasks completed within 30 seconds. Composer also supports voice-to-code generation, automated testing and debugging of code with native browser tools, and introduces a multi-agent collaborative architecture that enables up to eight agents to run in parallel.
