OpenAI Realtime GA + Realtime-mini tier: production-grade voice agents cross the always-on threshold

2 sources

OpenAI's GPT Realtime model page documents the GA of the production voice tier (server-side VAD, mid-session system-instruction updates, function calling during live sessions) and the launch of a Realtime-mini tier targeted at always-on voice agents. With GA, voice-agent creators can ship production traffic against a stable API; the Realtime-mini tier is positioned for high-volume / always-on assistants where cost-per-conversation dominates economics. The combination of GA stability, mid-session system-instruction updates, and a low-cost tier creates the conditions for voice agents to move from demo to product.

Why now

OpenAI Realtime GA is the inflection point where voice agents stop being 'cool demos that burn credits' and start being 'product features that ship in production'. The Realtime-mini tier lets creators build always-on assistants without the cost anxiety of earlier preview pricing.

Why it is worth publishing

Demo potential is wide: side-by-side conversation latency on a low-cost tier vs a standard tier, function-calling during a live session (e.g., ordering, scheduling, RAG), and a cost calculator showing the break-even volume where the Realtime-mini tier wins.

Evidence basis

OpenAI platform docs (high credibility) + The Decoder weekly roundup + IT之家 weekly roundup + cross-listed across model and tooling categories in the week's news cycle

“OpenAI Realtime just hit GA — and the Realtime-mini tier is what lets you ship an always-on voice assistant without watching the cost meter burn through your runway.”

Angle

Treat the Realtime GA + Realtime-mini tier as the 'voice agents ship to production' moment — frame the always-on tier as the unlock for 24/7 assistants.

Format

Short talking-head video

Demo idea

Record a 6-minute video: 90 seconds on 'what GA changes for production traffic', 90 seconds on 'function calling during a live session (book a flight by talking to the assistant)', 90 seconds on 'the Realtime-mini tier math (always-on assistant break-even)', 90 seconds on 'where to NOT use Realtime (high-stakes compliance, low-latency streaming)'.

Platform notes

Per-tier pricing beyond the documented audio-token rates is not extracted in this pass; do not state specific dollar-per-token figures. Vendor framing of GA is the vendor's talking point — confirm against your own test calls.

Usable claims

OpenAI Realtime reached general availability for production voice-agent traffic in the captured model page, with a Realtime-mini tier aimed at always-on voice assistants and function-calling during live sessions.

Evidence pipeline

From the news

Breakdown

OpenAI Realtime GA + Realtime-mini tier is the moment voice agents ship to production — but the demo potential is so wide that creators can easily lose focus on the 'always-on cost math' story. This explainer frames the production voice surface (server-side VAD, mid-session system-instruction updates, function calling during live sessions) and the always-on tier (Realtime-mini) as a single coherent 'production ship' story.

Sources

Risks

Model page documents the existence of the Realtime-mini tier but the precise per-tier rate beyond the captured summary is not in this pass. Verify specific capability claim against the underlying vendor docs and the actual license / pricing matrix before stating it on the record; do not paraphrase per-platform pricing or license terms into specific dollar figures or commercial-use clauses.
Each release-notes page is the vendor's primary source and frames its release against the competitive set the vendor cares about. Use The Decoder and IT之家 as media-type corroboration, but read the underlying vendor docs for any specific capability claim before stating it on the record. Verify specific capability claim against the underlying vendor docs and the actual license / pricing matrix before stating it on the record; do not paraphrase per-platform pricing or license terms into specific dollar figures or commercial-use clauses.

Demo ideas

Run a side-by-side latency demo on the standard tier vs the Realtime-mini tier — 10 turns each, plot the distribution.
Build a function-calling live demo: ask the assistant to book a calendar slot by voice; show the function-call JSON in a side panel.
Cost calculator: estimate break-even volume where the Realtime-mini tier wins for an always-on customer-support assistant.