Cursor 3 replaces the traditional code editor with an AI‑agent‑centric console, signaling a major shift in developer workflows, pricing, and market dynamics as AI‑driven tooling challenges the dominance of VS Code and other classic IDEs.
Cursor’s new Composer 2 model outperforms Claude Opus 4.6 on benchmarks like Terminal‑Bench 2.0, slashes pricing to $0.5/2.5 USD per million tokens, and introduces a self‑summary reinforcement‑learning technique that dramatically reduces context loss in long‑running coding tasks.
Cursor unveiled Composer 2, a self‑trained code‑only LLM that scores 61.3 on its CursorBench, outperforms Claude Opus 4.6 and rivals GPT‑5.4, while charging just $0.50 per million tokens—about one‑tenth the price of competitors—prompting a new AI pricing battle.
The article examines Cursor’s Composer 2 model, detailing its self‑summarization reinforcement‑learning workflow, the limitations of traditional compression methods, token‑efficient results on the CursorBench benchmark, and a challenging Terminal‑Bench case study that demonstrates dramatically reduced token usage while improving performance.
A developer uncovered that Cursor’s Composer 2 actually runs on the Kimi K2.5 model with reinforcement learning, prompting a rapid licensing dispute that ended with official confirmation and highlights the opaque yet collaborative nature of today’s open AI model ecosystem.
Cursor’s newly released Composer 2 model surpasses Claude Opus 4.6 on benchmarks such as Terminal‑Bench 2.0, offers dramatically lower token pricing, and achieves these gains by introducing a novel self‑summary reinforcement‑learning technique that compresses long‑context tasks while preserving critical information.
Cursor's Composer 2, a code‑only AI model, jumps from a 44.2 to 61.3 benchmark score, outperforms Claude Opus 4.6, nears GPT‑5.4, and costs just $0.50 per million tokens, reshaping its strategy after heavy reliance on external APIs.
