What is the issue?
Using ollama 0.30.6 with gemma4:latest (c6eb396dbd59) on Windows the model seems unable to see image files:
ollama run gemma4:latest what is this image ./cat.jpg
Added image './cat.jpg'
Thinking...
Thinking Process:
1. **Analyze the Request:** The user has provided an image and asked, "what is this image."
2. **Examine the Image:** I need to look closely at the input image.
3. **Initial Observation (Self-Correction/Constraint Check):** My capabilities include processing text and
images. However, in a typical LLM interaction where I am responding based on an *uploaded* or *provided* visual
asset, if the user simply pastes a corrupted link, placeholder text, or an uninterpretable character string (like
the "waj" at the end of the provided snippet), I must assess what information is actually present.
4. **Analyze the Provided Visual Data:** The image data provided by the user appears to be: `what is this
image\n[Image Placeholder/Corruption]`. Crucially, the actual visual content seems to be highly abstract or
potentially corrupted during transmission/display in the prompt structure (it looks like a block of text with
strange characters and formatting).
5. **Identify the Core Problem:** I cannot clearly interpret the image due to its nature in the provided context.
It is not a standard, recognizable photograph, drawing, or object. The visible characters suggest illegibility or
damage.
6. **Formulate the Response Strategy:**
* Acknowledge that an image was provided.
* State that I am having trouble interpreting the specific content (due to low resolution, corruption, or
abstract nature).
* Ask the user for clarification or more context if possible.
7. **Drafting the Response (Internal Monologue Check):** *The image is unclear. I must tell the user I need a
better image.*
8. **Final Output Generation.** (This leads to the suggested polite and helpful response.)
...done thinking.
I cannot clearly identify what this image depicts because the picture you provided is very abstract, blurry, or
potentially corrupted/too dark for me to make out specific details.
Could you please:
1. **Try uploading a clearer version of the image?**
2. **Provide more context about where this image came from?** (e.g., "This came from an antique book," or "This is
supposed to be a map.")
Once I have a better view, I would be happy to tell you what it is!
The GUI has the same issue, it let's you add the image but the model cannot see it.
With gemma3 it works:
ollama run gemma3:latest what is this image ./cat.jpg
Added image './cat.jpg'
The image shows a beautiful, fluffy **British Shorthair cat**. It's a dark grey (blue) color with striking amber
eyes. The cat is sitting in a relaxed posture.
Any idea on what might be the problem?
Relevant log output
OS
Windows
GPU
Nvidia
CPU
Intel
Ollama version
0.30.6
What is the issue?
Using ollama 0.30.6 with gemma4:latest (c6eb396dbd59) on Windows the model seems unable to see image files:
The GUI has the same issue, it let's you add the image but the model cannot see it.
With gemma3 it works:
Any idea on what might be the problem?
Relevant log output
OS
Windows
GPU
Nvidia
CPU
Intel
Ollama version
0.30.6