I’ve used this feature around 2-3 months ago, i’m wondering if image generation is still available with request based plans. I tried using Compose 2.5, Auto and Opus 4.6 in agent mode and all were refusing.
Hey, thanks for the question. Image generation on the legacy request-based plan is still available, but only with Max Mode turned on. That’s expected on request-based plans, and the same applies to subagents and the newest frontier models.
How to enable it:
- Open a chat in Agent mode and open the model picker.
- Pick a model that has the Max Mode toggle, for example Claude Opus 4.6. Auto and Composer 2.5 don’t have it, which is why they refused.
- Turn on Max Mode.
- Ask for an image explicitly, like “Generate an image of …”.
Note: With Max Mode on, this chat is billed at the model’s API rate plus 20%, not as a request. The generated image is billed the same way.
If this used to work without Max Mode, it’s because the Max Mode requirement wasn’t applied correctly for everyone for a while. That was fixed in early July, so Max Mode is required now.
More details on Max Mode for legacy plans: Models & Pricing | Cursor Docs
Let me know if it still doesn’t work after turning on Max Mode.