[Feature] Add gpt-4o-mini pricing for Azure OpenAI

[Manouchehri]

          Please checkout Azure GPT-4o-mini ;) https://techcommunity.microsoft.com/t5/ai-azure-ai-services-blog/openai-s-gpt-4o-mini-now-available-in-api-with-vision/ba-p/4200640

Originally posted by @seonghobae in https://github.com/BerriAI/litellm/issues/4922#issuecomment-2262822864

[krrishdholakia]

this is live already @Manouchehri https://github.com/BerriAI/litellm/blob/2a95484a83a2d16ae3180b5459ecd204c537b9ae/model_prices_and_context_window.json#L499

[krrishdholakia]

any reason you thought it wasn't supported yet?

asking to improve here

[Manouchehri]

I was expecting azure/gpt-4o-mini-2024-07-18, not azure/gpt-4o-mini. =)

[Manouchehri]

@krrishdholakia Am I making an obvious mistake here...? It seems like the cost is always zero, even after changing to azure/gpt-4o-mini.

[Manouchehri]

model_list:
  - model_name: gpt-4o-mini-2024-07-18
    litellm_params:
      model: azure/gpt-4o-mini-2024-07-18
      api_version: "2024-06-01"
      azure_ad_token: "oidc/google/https://example.invalid"
      additional_drop_params: ["user"]
      api_base: "https://gateway.ai.cloudflare.com/v1/REMOVED/litellm/azure-openai/REMOVED"
      seed: 1337
      max_tokens: 4096
      logprobs: True
      top_logprobs: 5
      frequency_penalty: 0
      presence_penalty: 0
      n: 1
      temperature: 1
      top_p: 1
      extra_headers: {"cf-skip-cache": "true"}
    model_info:
      base_model: azure/gpt-4o-mini

[krrishdholakia]

Nope. Works just fine. Tested on braintrust + langfuse + checked the response cost on curl

If you're seeing issues with cost calc - i'd recommend bumping version, and if the issue persists create a new issue w/ complete server logs so we can debug.

[krrishdholakia]

Here's my config for repro -

model_list:
  - model_name: "gpt-4o-mini"
    litellm_params:
      model: "azure/my-gpt-4o-mini"
      api_base: os.environ/AZURE_API_BASE
      api_key: os.environ/AZURE_API_KEY
    model_info:
      base_model: azure/gpt-4o-mini

litellm_settings:
  success_callback: ["langfuse"]

[krrishdholakia]

I was expecting azure/gpt-4o-mini-2024-07-18, not azure/gpt-4o-mini. =)

Any reason why? I can see we have this behaviour in some places of the model cost map, but in general we try to adhere to what the model name that is returned from the provider would look like