GPT-Realtime-2.1

Tool callingReasoningImage inputAudio inputAudio output

Website

Realtime speech-to-speech model with configurable reasoning, tool use, and robust voice-agent behavior

Released: 3mo ago
Pricing: Not published

Sources:models.dev

Benchmarks

DeepSWE is the primary coding benchmark where an exact model evaluation exists. Other sourced evaluations appear below; missing scores are not zero.

Not evaluated in DeepSWE 1.1 as of 2026-09-22. Results are only shown for exact model versions in the official dataset. Official leaderboard.

No benchmark results have been matched to this model yet.

Specifications

Developer: OpenAI
Model family: GPT
Type: Audio model
Predecessor: —
Successor: —
Released: July 6, 2026
Model ID: openai/gpt-realtime-2.1
Openness: Closed
License: —
Input modalities: text, audio, image
Output modalities: text, audio
Modality: text+audio+image->text+audio
Hugging Face: —
Source freshness: Metadata synced Oct 8, 2026, 10:17 PM UTC

Additional source evidence

Source model: openai/gpt-realtime-2.1
Source family: gpt
Last updated: 2026-07-06
Knowledge cutoff: 2024-09-30
Model context: 128K
Maximum input: 96K
Model output limit: 32K
Open weights: No
File attachments: Yes
Reasoning: Yes
Tool calling: Yes
Structured output: No
Temperature control: No

Complete source record (JSON)

{
  "id": "openai/gpt-realtime-2.1",
  "name": "GPT-Realtime-2.1",
  "description": "Realtime speech-to-speech model with configurable reasoning, tool use, and robust voice-agent behavior",
  "family": "gpt",
  "last_updated": "2026-07-06",
  "knowledge": "2024-09-30",
  "attachment": true,
  "reasoning": true,
  "tool_call": true,
  "structured_output": false,
  "temperature": false,
  "modalities": {
    "input": [
      "text",
      "audio",
      "image"
    ],
    "output": [
      "text",
      "audio"
    ]
  },
  "limit": {
    "context": 128000,
    "input": 96000,
    "output": 32000
  },
  "release_date": "2026-07-06",
  "open_weights": false
}

Infrastructure

Provider availability

All providers

Provider-specific media pricing is not available in the imported source data yet.

Text provenance

EU · product-scoped

OpenAI textGrain watermarking

In the EU, ChatGPT-generated text uses textGrain. OpenAI says API customers can opt in globally for select models, so watermarking depends on the product surface and configuration.

Official explanation