Gemini-1.5-flash-001-tuning is not found for API version v1beta

Gemini API
,
1

Starting today, we are receiving this error for our tuned model completions using the Python google-genai package. The tuned models are visible when listing them via client.tunings.list().

google.genai.errors.ClientError: 404 NOT_FOUND. {'error': {'code': 404, 'message': 'models/gemini-1.5-flash-001-tuning is not found for API version v1beta, or is not supported for generateContent. Call ListModels to see the list of available models and their supported methods.', 'status': 'NOT_FOUND'}}

Here are the tuned model details. The supported_actions field value is questionable.

version: None
endpoints: None
labels: None
tuned_model_info: base_model='models/gemini-1.5-flash-001-tuning' create_time=datetime.datetime(2025, 4, 10, 22, 57, 50, 445049, tzinfo=TzInfo(UTC)) update_time=datetime.datetime(2025, 4, 11, 2, 39, 7, 969044, tzinfo=TzInfo(UTC))
input_token_limit: None
output_token_limit: None
supported_actions: None

Any info would be highly appreciated.

2

Welcome to the forum!

For fine-tuning I would highly recommend considering Vertex AI.

Regards

3

Is the non-Vertex API being deprecated? Our project is currently down due to this…

4

Model tuning is not directly supported via the API now. For fine-tuning, I recommend using Vertex AI.

5

Still, the question is why the non-Vertex tuned model inference is down. What has changed, and is this permanent?

7

Hi pranciskus,
You’re encountering this error:
models/gemini-1.5-flash-001-tuning is not found for API version v1beta, or is not supported for generate Content

This indicates that your tuned model exists, but the generateContent method is not supported for it in the v1beta API or the model name is not being accessed correctly.

Verify the tuned model’s full name

Use client.tunings.list() and inspect the full resource name of the model. It might look like:
projects/your-project-id/locations/us-central1/tunedModels/your-model-id

9

Hi @Dinesh_Kumar79! The model name is correct. I see it when running client.tunings.list(). The same code was working just fine yesterday.

10

When creating a new finetune job, we are getting this error: Error creating tuned model: 400 models/gemini-1.5-flash-001-tuning is not found for CREATE TUNED MODEL at API version v1beta.. It appears that the entire finetuning functionality was silently removed from the API, including finetuned model inference. No communication about this surely does not promote confidence. I still hope this is a temporary issue and we will have a grace period for the transition.

12

To add to the current issues with model tuning and accessing already tuned models:

  1. As of today, none of the models currently available via the google.genai or google.generativeai APIs support the createTunedModel method. Any model that was previously finetuned by the user (derived from finetuneable Gemini versions such as gemini-1.5-flash-001-tuning model) cannot be called, despite still being listed by google.genai.Client.tunings.list() or google.generativeai.list_tuned_models().
  2. The documentation found at https://ai.google.dev/gemini-api/docs/model-tuning/tutorial?lang=python seems to be outdated.
    The finetuning instructions reference the gemini-1.5-flash-001-tuning model, which is no longer accessible via either the current (google.genai) or deprecated (google.generativeai) APIs. The notebooks linked on the page also do not work.

Summary:

  • If the removal of access to finetuneable models was intentional, this change should be announced as it breaks existing workflows.

  • The inability to call existing models finetuned by the user could potentially be a side effect, likely because their base models (gemini-1.5-flash-001-tuning model) are no longer accessible.

  • Finetuning instructions are outdated.

1 Like