Qwen3 fails to read uploaded images, only receives filename in chat

[jacklin719772]

What happened?

When uploading an image in the LibreChat chat window, the Qwen multimodal model cannot read the actual image content. The model only receives the filename or a placeholder, preventing it from analyzing the image or answering related questions. the result is that after uploading an image, the model cannot analyze or respond to questions about the image, causing multimodal features to fail.

Version Information

ghcr.io/danny-avila/librechat-dev latest 005b44167f14 3 days ago 1.57GB
ghcr.io/danny-avila/librechat-rag-api-dev latest df907f9efe75 10 days ago 9.61GB
ghcr.io/danny-avila/librechat latest fdaddca5d75c 12 days ago 1.57GB

Steps to Reproduce

On the chat window:

1. Click Upload an Image, then select a JPEG file from my local machine and upload it to the server.
2. After the image is uploaded successfully, enter a question below the image thumbnail, for example:"What is in the image?"
3. Click the Send icon.
4. The model starts processing (thinking).
5. The model only provides the above feedback (shows a placeholder or filename) and cannot answer the question.

What browsers are you seeing the problem on?

Chrome

Relevant log output

no error log display

Screenshots

Code of Conduct

• I agree to follow this project's Code of Conduct

[danny-avila]

According to Ollama model page, this model does not support vision so it may be getting excluded at that layer:
https://ollama.com/library/qwen3:30b-a3b