Prompt token count of exceeds the limit of 128000

[DanielProkhorov17]

Description

I get this error when a model like claude is trying to read file attachments that are images (.png, jpeg), thus crashing with token count limit.

I'd be great to perform token count before the request is getting sent to the model. In case, it exceeds the images shall be compressed to a reasonable size.

Plugins

No response

OpenCode version

1.1.48

Steps to reproduce

1. Let the model download an image (for e.g. via MPC), or python script ivocation, or curl
2. The AI model tries to read downloaded image(s)

Screenshot and/or share link

No response

Operating System

Windows 11

Terminal

Windows Terminal

[github-actions]

This issue might be a duplicate of existing issues. Please check:

• ContextWindowExceededError during compaction (litellm & claude-opus-4.5) #11450: ContextWindowExceededError during compaction (litellm & claude-opus-4.5) - Similar context/token limit issues
• Read tool fails to read image files even with vision-capable models (paste works) #11306: Read tool fails to read image files even with vision-capable models - Related to image file handling causing context issues
• Read tool accesses JPEG file but vision model hallucinates incorrect description (no actual image content processed) #11437: Read tool accesses JPEG file but vision model hallucinates incorrect description - Related to image processing and model context limits
• Images returned from Read tool don't work with OpenAI-compatible providers #11304: Images returned from Read tool don't work with OpenAI-compatible providers - Related to image handling in Read tool
• Read tool for kimi k2.5 on zen dose not work on images #11550: Read tool for kimi k2.5 on zen does not work on images - Similar image handling issue
• Model context limits not respected #11286: Model context limits not respected - Related to context window limit management

Feel free to ignore if none of these address your specific case.

[DanielProkhorov17]

#11450 This issue is reported during compaction. The usecase I'm describing is specifically related to image attachments.

[bankjaneo]

The last good version that can read image is 1.1.31

You might want to pin to that version until this bug is fixed.

[tlaukkanen]

Thanks for the tip. 1.1.31 works with images. Especially handy when agent uses MCP like playwright to troubleshoot some bug and tries to take a screenshot.

[thomashfr]

I'm getting the same error when using the Figma prototype, or any more detailed prompt in plan mode. Version 1.1.48

[paq85]

Happens also with local model, when it tries to analyse image

[aeciolevy]

1.1.31

Lifesaver.

I could not do anything with images anymore.

[bankjaneo]

This bug has been fixed in 1.1.52 🎉