When will the new GPT-5.6 Responses API parameters be supported in AIP?

Ask the Community
1

Hi Palantir team!

Thanks for making the GPT-5.6 series available in AIP.

We’ve started testing the models through AIP’s OpenAI-compatible Responses endpoint:

POST /api/v2/llm/proxy/openai/v1/responses

However, some of the newer GPT-5.6 parameters do not appear to be supported yet. For example, setting reasoning.effort to "max" results in a “Failed to deserialize request body” error.

Is support planned for the following parameters?

  • reasoning.effort: "max"
  • reasoning.mode
  • service_tier
  • Other new GPT-5.6 Responses API capabilities

For service_tier specifically, will options such as default and priority be exposed through the AIP endpoint, or will service-tier selection be managed internally by AIP?

Are these parameters temporarily unavailable because the proxy schema has not yet been updated for GPT-5.6 and the latest Responses API parameters, or are any of them intentionally unsupported?

Thanks!