Get API access
Model release

GPT Realtime Translate: Scope and Evaluation

OpenAI released GPT Realtime Translate on May 7, 2026. This brief outlines the confirmed API endpoints, model roles, and a practical testing procedure for…

Released Published 3 min readModel evidence ↗
GPT Realtime Translate release illustration — OpenAI

Release Details and Model Roles

The source event occurred on May 7, 2026. On this date, OpenAI released three models: GPT-Realtime-2, GPT-Realtime-Translate, and GPT-Realtime-Whisper. GPT-Realtime-2 is described as a realtime voice model with configurable reasoning for speech-to-speech agents. GPT-Realtime-Translate is designated for streaming speech translation. GPT-Realtime-Whisper is intended for streaming speech-to-text tasks.

  • GPT-Realtime-2 supports configurable reasoning for speech-to-speech agents.
  • GPT-Realtime-Translate handles streaming speech translation.
  • GPT-Realtime-Whisper performs streaming speech-to-text conversion.

API Endpoints and Documentation Updates

The release lists three API endpoints: v1/realtime, v1/realtime/translations, and v1/realtime/transcriptionsessions. The v1/realtime endpoint is associated with general realtime functionality. The v1/realtime/translations endpoint corresponds to translation capabilities. The v1/realtime/transcriptionsessions endpoint supports transcription sessions. Documentation updates include an updated Realtime and audio guide, a dedicated Realtime translation guide, and refreshed Realtime transcription documentation for streaming transcripts. Realtime prompting guidance was also moved within the documentation structure.

  • v1/realtime endpoint is listed for general use.
  • v1/realtime/translations endpoint supports translation workflows.
  • v1/realtime/transcriptionsessions endpoint manages transcription sessions.

Evaluation Procedure and Evidence Limits

Engineers can evaluate these models by designing controlled tests. First, test the v1/realtime/translations endpoint with a standardized set of audio inputs in multiple languages. Measure the latency between audio input and translated output. Compare the accuracy of the translations against a reference dataset. Validate the stability of the streaming connection over extended periods. The source does not provide benchmark scores, pricing information, or context window limits. The openWeights field is false, which records that open weights are not confirmed; it does not establish that local installation is impossible or that API access is mandatory. No architectural details or deployment requirements are specified in the source material.

  • Test translation accuracy using a fixed multilingual audio corpus.
  • Measure end-to-end latency for streaming responses.
  • Validate session persistence during long-duration streams.

Primary sources

QUICK ANSWERS

Frequently asked questions

What specific models were released on May 7, 2026?

OpenAI released GPT-Realtime-2, GPT-Realtime-Translate, and GPT-Realtime-Whisper. GPT-Realtime-2 focuses on speech-to-speech agents with configurable reasoning, while the other two handle streaming translation and transcription respectively.

Does the source confirm if these models can be installed locally?

No. The source states that open weights are not confirmed (openWeights=false). This fact does not establish whether local installation is possible or impossible, nor does it mandate API-only usage. The source does not provide deployment architecture details.