Repository navigation
[Bug] look_at tool cannot read image files - "this model does not support image input" even with vision-capable model #16184
Description
Bug Description
The look_at tool fails to analyze image files from disk, returning error "Cannot read image (this model does not support image input)" even though the model supports multimodal vision capabilities.
Steps to Reproduce
- Start OpenCode with vision-capable model (qwen3.5-397b or similar)
- Call
look_attool with image file path:
look_at(file_path="/Users/myname/Desktop/ScreenShot_2026-03-05_215547_277.png", goal="Analyze this screenshot") - Observe error: "Cannot read [filename] (this model does not support image input)"
Expected Behavior
The image should be passed to the vision-capable model for analysis via the look_at tool.
Actual Behavior
look_attool fails with "model does not support image input" error- Model never receives the image data
- However, pasting/dragging the same image directly into chat works perfectly
Environment
- OpenCode version: [请运行
opencode --version获取版本号] - OS: macOS
- Model: qwen3.5-397b (vision-capable)
Related Issues
- Read tool cannot pass image data to vision-capable models #15728 - Read tool cannot pass image data to vision-capable models
- [ModelScope] Custom Provider with Qwen Model Cannot Recognize Images - "Image Input Not Supported" #15040 - ModelScope custom provider with Qwen model cannot recognize images
- Read tool fails to read image files even with vision-capable models (paste works) #11306 - Read tool fails to read image files (paste works)
Additional Context
This appears to be a core issue with the tool chain failing to convert image files into proper visual input format for vision-capable models. Pasted images work fine, but file-based image reading fails.
opencode --version
1.2.17
Activity
github-actions commentedon Mar 5, 2026
This issue might be a duplicate of existing issues. Please check:
If your issue is distinct from the above, please add a comment explaining how it differs.
mtomas7 commentedon Mar 10, 2026
@Ryker0716 I am trying to resolve a similar issue with new Qwen3.5 model with Vision, but as per OpenCode docs there is not such look_at tool. Is this a custom-built tool?
Ryker0716 commentedon Mar 10, 2026
The problem occurred because I used the oh-my-opencode plugin. It works normally after disabling the plugin.
mtomas7 commentedon Mar 10, 2026
Please see if this config tip will help: #15728 (comment)
roadmap: #220 filed — Image/vision input is structurally impossible a…
roadmap: #220 filed — Image/vision input is structurally impossible a…
roadmap: #220 filed — Image/vision input is structurally impossible a…
roadmap: #220 filed — Image/vision input is structurally impossible a…
roadmap: #220 filed — Image/vision input is structurally impossible a…
roadmap: #220 filed — Image/vision input is structurally impossible a…
roadmap: #220 filed — Image/vision input is structurally impossible a…
roadmap: #220 filed — Image/vision input is structurally impossible a…