Lowest price
Choose GPT-Realtime Mini
GPT-Realtime Mini costs $10.00 / 1M audio tokens, against $32.00 / 1M audio tokens for GPT-Realtime-2.1.
Choose GPT-Realtime-2.1 for very long inputs, the latest release and the most recent knowledge. Choose GPT-Realtime Mini for the lowest price.
Each answer below comes straight from the published prices, specifications and independent benchmark results.
Choose GPT-Realtime Mini
GPT-Realtime Mini costs $10.00 / 1M audio tokens, against $32.00 / 1M audio tokens for GPT-Realtime-2.1.
Choose GPT-Realtime-2.1
GPT-Realtime-2.1 reads up to 128k tokens per request, against 32k for GPT-Realtime Mini.
Choose GPT-Realtime-2.1
GPT-Realtime-2.1 is the newer release (Jul 6, 2026, against Oct 6, 2025 for GPT-Realtime Mini).
Choose GPT-Realtime-2.1
GPT-Realtime-2.1 was trained on data up to September 2024, against October 2023 for GPT-Realtime Mini.
| GPT-Realtime-2.1 | GPT-Realtime Mini | |
|---|---|---|
| Provider | OpenAI | OpenAI |
| Category | Voice and realtime | Voice and realtime |
| Status | Generally available | Deprecated |
| Headline price | $32.00 / 1M audio tokens | $10.00 / 1M audio tokens |
| Context window | 128k | 32k |
| Max output | 32k | 4k |
| Released | Jul 6, 2026 | Oct 6, 2025 |
| Knowledge cutoff | September 2024 | October 2023 |
| Licence | Proprietary | Proprietary |
| Capabilities | Audio input, Reasoning, Tool calling | Audio input, Tool calling |
| API model id | gpt-realtime-2.1 | gpt-realtime-mini |
Choose GPT-Realtime-2.1 for very long inputs, the latest release and the most recent knowledge. Choose GPT-Realtime Mini for the lowest price.
GPT-Realtime Mini is 69% cheaper than GPT-Realtime-2.1 at the headline price ($10.00 / 1M audio tokens vs $32.00 / 1M audio tokens).
GPT-Realtime-2.1 reads more per request: 128k tokens against 32k.
GPT-Realtime-2.1 is made by OpenAI and GPT-Realtime Mini by OpenAI. GPT-Realtime Mini is 69% cheaper than GPT-Realtime-2.1 at the headline price ($10.00 / 1M audio tokens vs $32.00 / 1M audio tokens). GPT-Realtime-2.1 reads more per request: 128k tokens against 32k. GPT-Realtime-2.1 is the newer model (Jul 6, 2026, against Oct 6, 2025). Only GPT-Realtime-2.1 offers reasoning.
GPT-Realtime-2.1 was released on Jul 6, 2026 and GPT-Realtime Mini on Oct 6, 2025.