Are AI Studio's 1.0-pro-vision-001 and API's gemini-1.0-pro-vision-latest exactly the same model ?

[Yuma]

With Vertex AI Studio, I get the response as expected, but via the API, the quality of the response may be lower, or it may give a wrong response different from AI Studio. I think I am using the same model (1.0-pro-vision-001) and parameters for the request, but is it different internally?

The image is an example in Japanese, but when I asked Vertex AI Studio to extract questions and answers for the same image, it extracted everything correctly, but when I used the API (postman), there are some incorrect parts extracted.
In postman, we use pro-vision-latest, which we recognize to mean the same as vertexAI's 001 model.

Why does the quality of the answers change?
Is it the same 001 model but different internally?
Thank  you in advance.

[lsolatorio]

Hi @Yuma, 

Thanks for reaching out! We're here to help you with your inquiries.

It sounds like you're concerned about the different results you're getting from the pro-vision models. While both models share the same core functionality, versioning can indeed play a role. The last 3 digits in the model name (1.0-pro-vision-001) indicate a stable version, which is likely why you're seeing better results with it. The 'latest' version (1.0-pro-vision-latest) may still be under development, with ongoing bug fixes and improvements that could cause slight variations in output for some prompts.

A stable version of a Gemini model does not change and continues to be available for approximately six months after the release date of the next version of the model.

You can identify the version of a stable model by the three-digit number that's appended to the model name. For example, gemini-1.0-pro-001 is version number one of the stable release of the Gemini 1.0 Pro model.

---

The latest version of a model is updated periodically and includes incremental updates and improvements. These changes might result in subtle differences in the output over time for a given prompt. The latest version of a model is not guaranteed to be stable.

---

I hope I was able to provide you with useful insights.