gpt-realtime-2
gpt-realtime-2 is OpenAI's most capable realtime voice model, bringing configurable reasoning effort, stronger instruction following, and more reliable tool use to speech-to-speech voice agents.
import { gateway } from '@ai-sdk/gateway';
export async function POST() { const { token, url } = await gateway.experimental_realtime.getToken({ model: 'openai/gpt-realtime-2', });
return Response.json({ token, url, tools: [] });}Playground
Try out gpt-realtime-2 by OpenAI. Usage is billed to your team at API rates. Free users (those who haven't made a payment) get $5 of credits every 30 days.
gpt-realtime-2Voice agent
Talk to a realtime agent. It listens to your voice and replies with audio.
Start the session and ask the agent something.
Providers
Route requests across multiple providers. Copy a provider slug to set your preference. Visit the docs for more info. Using a provider means you agree to their terms, listed under Legal.
| Provider |
|---|
OpenAI | $4.00/M+1 more | $24.00/M | Read:$0.4/M Write:— | $10.00/K + input costs | — | 05/07/2026 |
More models by OpenAI
| Model |
|---|
| 1M | 0.7s | 56tps | $5.00/M | $30.00/M | Read: $0.5/M Write: — | $10.00/K + input costs | — | +4 |   | 04/24/2026 | |||
| 400K | 1.1s | 137tps | $0.75/M | $4.50/M | Read:$0.07/M Write:— | $10.00/K + input costs | — | +4 | | 03/17/2026 | |||
| 1.1M | 3.0s | 84tps | $2.50/M | $15.00/M | Read: $0.25/M Write: — | $10.00/K + input costs | — | +4 | | 03/05/2026 | |||
| 400K | 3.5s | 119tps | $0.25/M | $2.00/M | Read:$0.03/M Write:— | $14/K + input costs | — | +4 | | 08/07/2025 | |||
| 131K | 0.2s | 651tps | $0.35/M | $0.75/M | Read:$0.25/M Write:— | — | — |   +5 | 08/05/2025 | ||||
| 128K | 0.6s | 71tps | $0.15/M | $0.60/M | Read:$0.07/M Write:— | $14/K + input costs | — | +3 | | 07/18/2024 |
About gpt-realtime-2
gpt-realtime-2 launched on May 7, 2026 as the successor to gpt-realtime-1.5 and OpenAI's most capable realtime voice model. The design goal: voice agents that reason through problems as conversations unfold, instead of only routing or reciting.
Reasoning effort is configurable across levels from minimal to xhigh, so you choose how much thinking each session gets. Low effort is the default and keeps responses quick. Higher effort raises answer quality on complex workflows, at the cost of latency and output tokens. When a caller interrupts mid-thought, gpt-realtime-2 discards the current chain of thought and starts a fresh turn.
Conversation mechanics improved alongside reasoning. Preambles let an agent acknowledge the request out loud while tools run, parallel tool calls fire multiple requests at once, and recovery behavior catches tool failures rather than freezing the call. OpenAI reports higher scores than GPT-Realtime-1.5 on audio reasoning and instruction-following benchmarks.
What To Consider When Choosing a Provider
- Configuration: Reasoning effort is the main tuning decision. Low is the default and keeps latency down for most production voice agents; higher settings improve reasoning at the cost of latency and output token usage. Realtime support on AI Gateway is in beta, with a WebSocket integration through the AI SDK's realtime hook and short-lived server-minted tokens.
- Zero Data Retention: AI Gateway does not currently support Zero Data Retention for this model. See the documentation for models that support ZDR.
- Authentication: AI Gateway authenticates requests using an API key or OIDC token. You do not need to manage provider credentials directly.
When to Use gpt-realtime-2
Best For
- Multi-step voice problem solving: Agents that reason through hard requests as the conversation unfolds
- Complex tool workflows: Parallel tool calls, lookups, and decisions handled mid-conversation
- Mixed-difficulty sessions: Reasoning effort settings match compute and latency to each turn
- Natural conversation under load: Preambles and recovery behavior keep calls flowing while tools run
Consider Alternatives When
- Latency-critical simple turns:
gpt-realtime-1.5keeps standard voice agents fast without reasoning overhead - High-volume voice features:
gpt-realtime-minidelivers realtime interaction at a cost built for scale - Transcription or speech only:
gpt-4o-transcribeandtts-1cover single-direction audio jobs
Conclusion
gpt-realtime-2 moves voice agents from routing calls to resolving them. Start at low reasoning effort, raise it only where conversations demand deeper thinking, and route everything through AI Gateway for unified auth, observability, and spend control.