Cursor Admits Its New Composer 2 Coding Model Was Built on Moonshot AI's Kimi K2.5
Cursor's newly launched Composer 2 — marketed as the company's most capable coding model yet — turns out to be built on top of Kimi K2.5, an open-source model from Chinese AI lab Moonshot AI. Community developers spotted the underlying model ID before Cursor officially confirmed it.
VP Lee Robinson acknowledged that roughly 25% of compute came from the Kimi K2.5 base model, with the rest from Cursor's own training. Composer 2 is priced at $0.50/$2.50 per 1M input/output tokens — an 86% reduction vs. its predecessor — and benchmarks above Claude Opus 4.6 on Terminal-Bench 2.0 while trailing GPT-5.4.
The disclosure highlights a growing trend: well-funded Western AI companies fine-tuning Chinese open-source models under their own branding. It also validates Kimi K2.5's quality as a frontier base model — and raises real questions about transparency in AI product launches. Expect this pattern to repeat as Chinese open-weight releases keep pace with closed frontier models.