Vibe Coding Platform Cursor Releases First In-House LLM, Composer, Promising 4X Speed Boost

Why It Matters

Composer’s speed and agentic design enable near‑real‑time AI‑assisted development, giving Cursor a competitive edge over tools like GitHub Copilot and signaling a shift toward models trained within live coding environments for more reliable enterprise adoption.

Summary

Cursor, the Vibe coding platform from Anysphere, unveiled Composer, its first in‑house coding large language model, as part of the Cursor 2.0 update. Composer is a reinforcement‑learned mixture‑of‑experts model that generates code at 250 tokens per second—about four times faster than comparable frontier systems—and completes most interactions in under 30 seconds while maintaining mid‑frontier intelligence. The model is already deployed by Cursor’s engineering team and is integrated into a multi‑agent environment that can run up to eight parallel agents, with enterprise‑grade controls and tiered pricing. Benchmarking on the internal "Cursor Bench" shows Composer matches the reasoning of leading models while delivering the highest recorded generation speed across all categories.