Why is Spring AI defaulting to gpt-4o instead of the configured gpt-3.5-turbo model?

Below is a summary of what’s happening and some steps you can take to troubleshoot the issue:

---

What’s Happening

Despite setting the model to “gpt-3.5-turbo” in your application properties, the error message shows a call to “gpt-4o”. This indicates that somewhere in the configuration or library defaults the model is falling back to “gpt-4o”. Since the API response is returning a 404 (with a clear message that “The model gpt-4o does not exist or you do not have access to it”), it means the request ultimately isn’t using the correct model as you intended.

---

Possible Causes

• Configuration Overrides:
  Check whether your project has multiple configurations (for example, additional property files or environment variable settings) that might be overriding your defined property. Even though you’ve set

  PROPERTIESCopied
  spring.ai.openai.model=gpt-3.5-turbo

  there might be an unintended override that causes the fallback to “gpt-4o” when the value isn’t injected correctly.

• Library Defaults or Updates:
  It’s possible that a newer version (or an internal misconfiguration) in the Spring AI library is defaulting to “gpt-4o” when no valid model property is present or if the provided value isn’t picked up. Sometimes library updates introduce a new default, so verify that you are using the version you expect or check the release notes for any breaking changes in model configuration.

• Fallback Logic in Code:
  The Spring AI project might have internal fallback logic that reassigns a default value if the provided model setting isn’t valid or recognized. If that logic mistakenly sets “gpt-4o” as the default, then you would encounter this error even if you intended to use “gpt-3.5-turbo”.

---

What You Can Do

1. Double-check Configuration Loading:
   Verify that the property spring.ai.openai.model is properly loaded. Look through all configuration sources (e.g., other properties/YAML files or environment variables) to ensure nothing is overriding your intended setting.

2. Examine Dependency Versions:
   Confirm you are using the expected version of the Spring AI library. If there is a change in a recent version, reviewing the release notes or migration guides might indicate if the default has changed.

3. Enable Debug Logging:
   Turn on debug logging for Spring configuration to track down which configuration source is setting the model property. This might reveal if your property isn’t being correctly picked up and if a default (“gpt-4o”) is being provided by the framework.

4. Review Customizations:
   Look into the project’s code (or linked repository) to see whether any component (or external configuration) uses a fallback value of “gpt-4o” when no model is explicitly configured.

5. Test with a Minimal Setup:
   Create a small test project or a minimal configuration that only sets spring.ai.openai.model=gpt-3.5-turbo and calls the OpenAI API. This helps isolate whether the problem lies in the property injection or elsewhere.

---

Final Note

If none of these steps help, consider raising an issue on the Spring AI project's GitHub repo. It’s possible that this behavior might be due to an internal bug or an undocumented default that needs to be clarified.

This advice is based on the behavior observed in the OpenAI API’s model configuration model and similar error cases encountered by others when the API is called with an unintended model identifier.