OpenAI adds GPT-Realtime-2.1-mini to the API with reasoning and tool use

OpenAI says GPT-Realtime-2.1-mini is now available in the API, and that it brings reasoning and tool use to OpenAI's Realtime mini lineup at the same cost as GPT-Realtime-mini (OpenAI Developers on X).

This matters if you build lower-cost voice or streaming agents on OpenAI's Realtime stack: the announcement positions reasoning and tool use as capabilities you can now reach at the mini price point, rather than only on larger Realtime models. As of publish time the specifics come from the developer announcement alone, so the claims below are limited to exactly what OpenAI stated.

Key Takeaways

• What: OpenAI says GPT-Realtime-2.1-mini is now available in the API (OpenAI Developers on X).
• What it adds: Reasoning and tool use for OpenAI's Realtime mini lineup (OpenAI Developers on X).
• Cost: OpenAI says it is available at the same cost as GPT-Realtime-mini (OpenAI Developers on X).
• Caveat: OpenAI's platform Models page and Realtime guide did not list the model at publish time, so treat anything beyond the announcement as unconfirmed.

What OpenAI Announced

The OpenAI Developers account states that "GPT-Realtime-2.1-mini is now available in the API, bringing reasoning and tool use to our Realtime mini lineup at the same cost as GPT-Realtime-mini" (OpenAI Developers on X).

That is the full scope of what OpenAI confirmed in the announcement: availability in the API, the addition of reasoning and tool use to the Realtime mini tier, and cost parity with GPT-Realtime-mini. OpenAI did not publish additional details on latency, supported modalities, API parameters, or migration steps in the post.

Why It Matters for Voice and Streaming Agents

If you are building real-time voice or low-latency streaming experiences, the value of a mini Realtime model is usually cost per interaction. OpenAI framing this release as reasoning and tool use "at the same cost as GPT-Realtime-mini" suggests you may be able to add those capabilities without moving up to a more expensive Realtime model, though you should validate fit against your own workload.

Because OpenAI's own docs had not yet listed the model at publish time, confirm the exact model identifier, pricing, and supported features against OpenAI's official Realtime documentation and pricing before you wire it into production.

What Is Not Yet Confirmed

At publish time, OpenAI's platform Models page and Realtime guide did not mention GPT-Realtime-2.1-mini. Documentation lagging a model's availability is common, but it means the following are not yet verifiable from a durable OpenAI source: specific pricing figures, exact API model string, supported audio or text modalities, context limits, and any performance or benchmark claims. We will update this article when OpenAI's platform docs or changelog list the model.