AI Model Watch
Models / OpenAI

GPT-Realtime-2

GA

Speech-to-speech realtime model. Input: text/audio/image; Output: text/audio. Prices listed are the TEXT-token rates ($4 in / $24 out / $0.40 cached). Audio tokens are billed separately at $32 input / $64 output per 1M; image input $5/1M. Max output 32K tokens.

GPT-Realtime-2 by OpenAI costs $4 per 1M input tokens and $24 per 1M output tokens ($9/1M blended), with a 128K (128.000-token) context window. It is generally available (GA), with a knowledge cutoff of 30 Sep 2024.

Last verified: 13 Aug 2026 · sourced from official provider documentation

Provider
OpenAI
Status
GA
Input price
$4 / 1M tokens
Output price
$24 / 1M tokens
Cached input
$0.4 / 1M tokens
Blended price
$9 / 1M tokens
Context window
128.000 tokens (128K)
Max output
32.000 tokens
Modality
text, audio, image
Open weights
No — API only
Knowledge cutoff
2024-09-30
Released
—
API string
gpt-realtime-2

Source: OpenAI official documentation ↗

See also

Explore more

All OpenAI models→

Prices, context windows and lifecycle status for every OpenAI model we track.

Compare models head-to-head→

Side-by-side price, context and status for the models developers weigh against each other.

FAQ

GPT-Realtime-2 — questions & answers

How much does GPT-Realtime-2 cost?

GPT-Realtime-2 is priced at $4 per 1M input tokens and $24 per 1M output tokens ($9/1M blended at a 3:1 input-to-output ratio), with cached input at $0.4 per 1M tokens.

What is the context window of GPT-Realtime-2?

GPT-Realtime-2 has a 128.000-token context window (128K), with up to 32.000 output tokens per request.

Is GPT-Realtime-2 deprecated?

No — GPT-Realtime-2 is generally available (GA) and not currently scheduled for deprecation or retirement in our tracker.

Does GPT-Realtime-2 support image or vision input?

Yes — GPT-Realtime-2 accepts image input. Listed modalities: text, audio, image.

Can I self-host GPT-Realtime-2?

No — GPT-Realtime-2 is proprietary. Its weights are not publicly released, so it is available only through OpenAI's API (or hosted partners), not for self-hosting.

What is the API model string for GPT-Realtime-2?

The API model identifier for GPT-Realtime-2 is "gpt-realtime-2" when calling OpenAI's API.

What is GPT-Realtime-2's knowledge cutoff?

GPT-Realtime-2's training knowledge cutoff is 30 Sep 2024.