Hi there, I am a team/enterprise plan user and we are still on the 500 request per month legacy plan. Recently composer-1.5 launched and we were excited to start using it.
However it was really disheartening to see that composer-1.5 consumes 2 requests instead of 1 like composer-1. It consumes the same as claude opus 4.* thinking which makes us not want to use this model - as it finishes up our request quota pretty fast even for simple questions.
Please provide some clarity on why this model consumes x2 requests than composer-1. It’s a major turnoff and we will prefer to use your competitors models instead as they consume lesser requests (1 for Gemini, GPT-5.3-*)
hi @valentinoPereira and thank you for your question as it is understandable that the difference is noticeable.
Composer 1.5 uses 2 premium requests per call because each invocation is effectively a much larger, more compute‑intensive model than Composer 1. It is designed to solve harder coding problems more reliably rather than just stream tokens cheaply. I would recommend reaching out to your enterprise contact at Cursor about pricing.
More details:
Composer 1.5 has undergone around 20× more RL fine‑tuning than Composer 1, with post‑training compute that exceeds the original pretraining, which is very expensive to run per query.
The model uses a thinking‑tokens mechanism, where it internally reasons for many more tokens on difficult tasks before answering, so a single “request” can involve a long hidden chain‑of‑thought pass instead of a short response.
It performs RL‑trained self‑summarization over long sessions, which means extra internal passes to compress and retain context so it can keep working accurately on large codebases.
Composer 1.5 is a high‑end agentic coding model beating Sonnet 4.5 on Terminal‑Bench 2.0.