Azure OpenAI: GPT Realtime API 2024-12-17: New Voices, Rate Limits, and o1 Model | SignalBreak

Azure OpenAI: GPT Realtime API 2024-12-17: New Voices, Rate Limits, and o1 Model | SignalBreak | SignalBreak

Source text

The gpt-4o-realtime-preview model version 2024-12-17 is available for global deployments in East US 2 and Sweden Central regions . Use the gpt-4o-realtime-preview version 2024-12-17 model instead of the gpt-4o-realtime-preview version 2024-10-01-preview model for real-time audio interactions. Added support for prompt caching with the gpt-4o-realtime-preview model. Added support for new voices. The gpt-4o-realtime-preview models now support the following voices: alloy , ash , ballad , coral , echo , sage , shimmer , verse . Rate limits are no longer based on connections per minute. Rate limiting is now based on RPM (requests per minute) and TPM (tokens per minute) for the gpt-4o-realtime-preview model. The rate limits for each gpt-4o-realtime-preview model deployment are 100 K TPM and 1 K RPM. During the preview, Azure AI Foundry portal and APIs might inaccurately show different rate limits. Even if you try to set a different rate limit, the actual rate limit is 100 K TPM and 1 K RPM. For more information, see the GPT real-time audio quickstart and the how-to guide . December 2024 o1 reasoning model released for limited access The latest o1 model is now available for API access and model deployment. Registration is required, and access will be granted based on Microsoft's eligibility criteria . Customers who previously applied and received access to o1-preview , don't need to reapply as they're automatically on the wait-list for the latest model. Request access: limited access model application To learn more about the advanced o1 series models see, getting started with o1 series reasoning models . Region availability Model Region o1 (Version: 2024-12-17) East US2 (Global Standard) Sweden Central (Global Standard) Preference fine-tuning (preview) Direct preference optimization (DPO) is a new alignment technique for large language models, designed to adjust model weights based on human preferences. Unlike reinforcement learning from human feedback (RLHF), DPO doesn't r

View original source

GPT Realtime API 2024-12-17: New Voices, Rate Limits, and o1 Model

More from Azure OpenAI

Get alerts for Azure OpenAI