Feature request for product/service
Cursor IDE
Describe the request
Even after sonnet 4.5, gpt-5 remains the best model for my type of usage by a significant margin (rust, solidity, node.js, some finance stuff). However, it’s a slow model, and gpt-5-codex doesn’t seem to achieve as good results as I get in the codex-cli - not fully sure if it’s a system prompt thing, or related to the adaptive reasoning effort used for Cursor’s integration (seems to be the case). So, I’d be extremely happy if:
- gpt-5-mini-fast and gpt-5-mini-high-fast were supported; It seems like the current medium reasoning version doesn’t use OpenAI’s priority lane, so its latency seems similar to gpt-5-fast’s. If a “fast” variant a high reasoning one are added, I think both would be really strong daily drivers for those like myself that love gpt-5 but sometimes need a faster model, yet smarter than sonnet for my use cases. From what I tested I still get really good and arguably stronger results with gpt-5-mini than sonnet.
- Add support/selection of the reasoning effort for gpt-5-codex, so its results are closer to what we get when using its CLI.
I love Cursor’s UX, specially the new Agent Window that lets me stay in flow while collaborating with gpt-5’s variants, but I still have to resort to the codex-cli when I have a bigger task and need to make sure the model stays in high reasoning effort. Those additions would make me really happy =D
Operating System (if it applies)
MacOS