Question about Edit/Update Model API endpoint

[harshjhajarvis-svg]

Describe the feature or potential improvement

Hi Langfuse Team,

I'm working with the Models API and need some clarification around updating existing models programmatically.

Current Situation

I can successfully create models using POST /api/public/models from the API reference documentation.

However, I now need to update existing model configurations such as:

• Tokenizer settings
• Input/output pricing
• Other model metadata as configurations evolve over time

Questions

1. Is there a dedicated PUT or PATCH endpoint available for updating an existing model via the API?
2. Should I reuse the same POST /api/public/models endpoint with an existing modelId to perform updates?
3. Which fields are editable after a model has been created?

Use Case

I'm managing custom models programmatically and need a way to update tokenizer configurations and pricing over time without deleting and recreating models.

Current Code Example

const response = await axiosClient.post(
  `${baseUrl}/api/public/models`,
  payload,
  {
    auth: {
      username: publicKey,
      password: secretKey,
    },
  }
);

Additional Context

• Langfuse version: v3.147.0 OSS
• Deployment type: Self-hosted

Any guidance on the recommended approach for updating existing models would be appreciated.

Thanks!

Additional information

No response

[jannikmaierhoefer]

At the moment, there is no public PUT or PATCH endpoint for updating an existing model via the Models API.

POST /api/public/models is create-only, so you should not reuse it with an existing modelId for updates. It does not accept modelId and will not update an existing model in place.

So today, model configs are not editable via the public API after creation. If you need to change tokenizer settings, pricing, or other model metadata, the current workaround is to delete the custom model and recreate it with the new configuration. For built-in models, the supported approach is to create a project-level override instead of modifying the built-in definition directly.

Internally, the UI uses a non-public upsert path that can update fields like matchPattern, tokenizerId, tokenizerConfig, and pricing tiers, but that functionality is not exposed through the public REST API.