Cursor's Composer 2.5 matches Opus 4.7 and GPT-5.5 benchmarks at a fraction of the cost

May 18, 2026

Cursor has shipped Composer 2.5, a major upgrade to its in-house AI coding model. It builds on the open-source Kimi K2.5 checkpoint from Moonshot and was trained on 25 times more synthetic tasks than its predecessor, Composer 2. Cursor says 85 percent of the compute budget went toward extra training and reinforcement learning.

On benchmarks like SWE-Bench Multilingual (79.8 percent) and CursorBench v3.1 (63.2 percent), Composer 2.5 matches Opus 4.7 and GPT-5.5. It's far cheaper per task, though: $0.50 per million input tokens and $2.50 per million output tokens, a fraction of what Anthropic and OpenAI charge. A faster variant with the same performance runs $3.00 and $15.00, respectively. Composer 2.5 is live in Cursor now. More details are in the model documentation.

Composer 2.5 matches Opus 4.7 and GPT-5.5 on CursorBench 3.1 but costs less than a dollar per task - compared to up to eleven dollars for the competition. | Image: Cursor

Cursor is already training a much larger successor model "from scratch" with SpaceX and xAI, using ten times the compute on the Colossus-2 cluster with one million H100 equivalents. SpaceX had previously announced plans to acquire Cursor for $60 billion.