Composer 1.5 - Improved reasoning over challenging coding tasks
Three months ago, @Cursor launched Composer 1, their first coding agent, and they just released a new update, introducing 1.5.
According to the blog announcement: [1]
Our new release, Composer 1.5, strikes a strong balance between speed and intelligence for daily use (...) Composer 1.5 is a thinking model. In the process of responding to queries, the model generates thinking tokens to reason about the user’s codebase and plan next steps. (...) To handle longer running tasks, Composer 1.5 has the ability to self-summarize. This allows the model to continue exploring for a solution even when it runs out of available context.
In a previous thread, Claude's (Opus, Sonnet) and Google's (Gemini) coding models were leading the way. [2]
Curious if this new release would attract more developers. Any experiences with Composer 1.5?
Replies
I really liked Composer 1 because of its speed and cost. When I have to get simple tasks done quickly I reach for it. This new model however seems super expensive. Apparently more expensive than Codex, which makes it a pretty bad value proposition. The direction seems valid though: quick models.
Humans in the Loop
spot on! in comparison, @OpenAI's GPT-5.3 Codex pricing (per 1M tokens):
Input + Cache Write: $1.75
Output: $14
Cache Read: $0.175
Humans in the Loop
interesting! i have mixed feelings on this tbh. is speed as important when products like @Warp and @Axel let you run multiple agents at once?
Migma AI
What about price? It's really expensive.
Humans in the Loop
good p! about Composer 1.5 pricing (per 1M tokens):
Input + Cache Write: $3.5
Output: $17.5
Cache Read: $0.35
full comparison + source: cursor.com/docs/models#model-pricing
Simplehuman
Tried it out for a whole day building something basic and I'm afraid I had to revert to Sonnet (which itself is not my primary model, Opus is). I consistently saw that there was code duplication and I had to remind and handhold Composer 1.5 to stay in the bounds and reuse existing systems.
Humans in the Loop
oh thanks for the feedback!
I tried Composer 1.5 yesterday. It was slower than Sonnet 4.5 and didn't seem to get the whole code context. That said i tried it for about 30 min of work so hard to judge.