Provides audio input
AI Models & Assistants / Knowledge library
GPT-Realtime-2.1
Reasoning model with tool use
Why learn GPT-Realtime-2.1?
Offers the documented OpenAI model capabilities through hosted.
The problem it solves
Provides model inference for supported text / audio / image inputs.
Where it came from
GPT-Realtime-2.1: Reasoning model with tool use Published aliases: gpt-realtime-2.1. Input: text, audio, image. Output: text, audio. Checked 2026-10-13. Model capabilities are distinct from tools available in a particular agent.
Build your understanding
Work through these exercises in order. Try the task before opening a reference.
- Explain what GPT-Realtime-2.1 does.
- Identify a use case for GPT-Realtime-2.1.
- Describe a limitation of GPT-Realtime-2.1.
- Name a compatible tool for GPT-Realtime-2.1.
- Compare GPT-Realtime-2.1 against an alternative.
Documentation and sources
- Primary source / model card (developers.openai.com)
Catalog research notes
- See the AI Workbench for source dates, pricing evidence and configuration details.
Connect the ideas
Built on
Real World Example
Automated Emergency Dispatch Voice Triage
Emergency medical dispatch centers handling high volumes of inbound distress calls.
The business problem
Human dispatchers struggle to maintain composure and accuracy during complex, multi-step triage protocols while simultaneously documenting patient symptoms in real-time.
How this topic solves it
The GPT-Realtime-2.1 model leverages its reasoning capabilities and tool use to conduct voice-based triage, guiding callers through standardized medical protocols while executing database lookups to verify patient history or local hospital capacity.
- Establish a persistent connection via the v1/realtime endpoint to enable low-latency, bidirectional audio streaming.
- Define custom function tools that allow the model to query external medical databases for patient records.
- Configure the model's reasoning effort parameter to balance the depth of clinical analysis against the need for immediate, conversational response.
- Implement interruption handling to allow the model to pause its current instruction if the caller provides urgent, clarifying information.
- Stream the final triage summary to the dispatch console via tool output, ensuring the human dispatcher receives structured data.
The flow shows how audio input is processed through the model's reasoning engine to trigger external tools.
Processes audio and reasoning
External tool for records
Receives structured summary
- Caller → streams audio → GPT-Realtime-2.1
- GPT-Realtime-2.1 → executes tool call → Medical DB
- Medical DB → returns data → GPT-Realtime-2.1
- GPT-Realtime-2.1 → sends summary → Dispatcher
Expected business outcome
Improved consistency in triage protocol adherence and reduced cognitive load on dispatchers, allowing them to focus on high-level coordination rather than manual data entry.
Limitations and checks
Configurable reasoning effort can introduce latency; if the model is set to high reasoning, the conversational flow may experience delays that are unsuitable for critical, time-sensitive emergency interactions. This can be checked by monitoring the time-to-first-token during high-stress simulation scenarios.