GPT-Realtime-2 benchmarks, pricing and specs
GPT-Realtime-2 is an OpenAI model released on May 7, 2026 that takes text, audio and image input and produces text and audio. It costs $4 per 1M input tokens and accepts up to 128,000 tokens of context. It scores 1079 Elo on Artificial Analysis Speech Arena, #10 of 11 current models (Artificial Analysis, measured October 13, 2026).
- Per 1M input tokens
- $4
- Context window
- 128K
- Artificial Analysis Speech Arena
- 1079 Elo · #10/11
Data verified October 13, 2026 · Dataset 2026-10-13
Key facts
- Provider
- OpenAI
- Model family
- GPT Realtime
- Release date
- May 7, 2026
- Status
- Current
- Access
- API only
- API model id
- gpt-realtime-2
- Context window
- 128,000 tokens
- Max output
- 32,000 tokens
- Input
- text, audio, image
- Output
- text, audio
- Task families
- audio
- Declared capabilities
- realtime voice, reasoning
GPT-Realtime-2 pricing
It is tied for the most expensive of 10 current audio models priced per 1M input tokens.
Independent benchmark results
Scores come from independent evaluators that ran the model themselves, pinned to a dated snapshot. Rank is among current models in the same snapshot. OpenAI's own launch figures are listed under strengths with the publisher named, because they are not measured the same way.
| Benchmark | Score | Rank | Measured | Source |
|---|---|---|---|---|
| Artificial Analysis Speech Arena | 1079 Elo | #10 of 11 | October 13, 2026 | Artificial Analysis · 2026-10-13 |
Strengths, cautions and best uses
Strengths
- OpenAI's most capable realtime voice model, with configurable reasoning for speech-to-speech agents
- Text, audio and image input in one Realtime session, with function calling and prompt caching
Cautions
- Available only on the Realtime endpoint, not Chat Completions, Responses or Batch
- Audio costs $32 input and $64 output per 1M tokens, eight and four times its text rates
- Higher reasoning effort adds latency and output tokens
Best for
- Speech-to-speech voice agents with tool calls
Alternatives to GPT-Realtime-2
The closest current audiomodels by price. The cheapest model that scores as well or better is listed first; failing that, the cheapest one within 3% of the benchmark's scale.
Cheaper, same or higher score
Gemini 3.8 Flash TTS
- 88% cheaper per 1M input tokens, $0.50 vs $4
- 195 points higher on Artificial Analysis Speech Arena, 1274 vs 1079
- From Google
Gemini 3.5 Live Translate Preview
- 13% cheaper per 1M input tokens, $3.50 vs $4
- From Google
GPT-4o Transcribe
- 38% cheaper per 1M input tokens, $2.50 vs $4
- 16K context window vs 128K
gpt-audio-1.5
- 38% cheaper per 1M input tokens, $2.50 vs $4
Compare GPT-Realtime-2 with
Other GPT Realtime models
GPT-Realtime-2 FAQ
How much does GPT-Realtime-2 cost?
GPT-Realtime-2 costs $4 per 1M input tokens at the standard tier. It is tied for the most expensive of 10 current audio models priced per 1M input tokens. Source: OpenAI, verified October 13, 2026.
What is the context window of GPT-Realtime-2?
GPT-Realtime-2 accepts up to 128,000 tokens of context. It can return up to 32,000 tokens in one response. That ranks #4 of 8 current audio models by context length.
How does GPT-Realtime-2 score on independent benchmarks?
GPT-Realtime-2 scores 1079 Elo on Artificial Analysis Speech Arena, #10 of 11 current models (Artificial Analysis, measured October 13, 2026). These are independent evaluations, not OpenAI's own launch figures.
Is GPT-Realtime-2 open source or API only?
GPT-Realtime-2 is available through OpenAI's API only; its weights are not published.
What are the best alternatives to GPT-Realtime-2?
The closest current alternatives by price and measured performance are Gemini 3.8 Flash TTS (88% cheaper per 1M input tokens, $0.50 vs $4), Gemini 3.5 Live Translate Preview (13% cheaper per 1M input tokens, $3.50 vs $4), GPT-4o Transcribe (38% cheaper per 1M input tokens, $2.50 vs $4) and gpt-audio-1.5 (38% cheaper per 1M input tokens, $2.50 vs $4).
Sources
- GPT-Realtime-2 model · OpenAI · official · verified October 13, 2026
- OpenAI API changelog · OpenAI · official · verified October 13, 2026
- Artificial Analysis Speech Arena leaderboard (text to speech) · Artificial Analysis · independent · verified October 13, 2026 · Artificial Analysis Terms of Use
Cite this page
Van Data Team. "GPT-Realtime-2 Benchmarks, Pricing and Specs." AI Model Benchmark Dataset 2026-10-13. https://vandatateam.com/ai-model-benchmarks/gpt-realtime-2
Reuse the data: JSON · CSV · or explore every model in the AI model benchmark comparison.
Production evaluation
Deciding whether GPT-Realtime-2 fits your workload?
Public benchmarks narrow the shortlist. We test the finalists on your own tasks, data and cost limits, then build the routing and review controls to run the winner in production.
- Evaluation cases and acceptance thresholds tied to real work
- Model routing, fallbacks and cost controls