On Thursday, Cursor, the AI-assisted code editor, took another shot at its much-larger competitors when it released Composer 2, an AI model that outperforms Anthropic’s Opus 4.6 on some key coding benchmarks, and does so at a fraction of the cost.
Composer 2 is the third generation of its in-house coding model in five months, writes Frederic Lardinois, Senior Editor for AI, showing that Anysphere, the company that makes Cursor, is playing at an allegro speed.
Previous models applied reinforcement learning to an existing base model without modifying the base itself, Lardinois writes. But Composer 2 is the first version where Cursor ran continuous pre-training, which the company says provided “a far stronger base to scale our reinforcement learning.”
A new training technique that teaches the model to summarize its own context is the key to Composer 2's performance leap.
Cursor on Thursday released Composer 2, the third generation of its in-house coding model. The model outperforms Anthropic’s Opus 4.6 on some key coding benchmarks, and does so at a fraction of the cost.
The new Cursor model costs as little as $0.5 per million input tokens and $2.5 per million output tokens. There is...
On-demand: Building a self-service developer platform the thoughtful way
Developers want a public cloud-like experience — even on-prem. In this insightful session, we uncovered how to automate full-stack Kubernetes deployments with a single Git commit. The entire webinar is now available on-demand so you can watch at your convenience.
Upcoming webinar: Building production-ready Agentic AI
Generative models alone aren’t enough for production-grade AI. Without a unifying control plane, scaling autonomous agents often leads to hallucinations, cascading errors, and unpredictable system behavior. Join us next week for a deep dive into the missing piece in the AI stack: the Control Plane.
