OpenAI Realtime API Alternatives Compared (2026)
OpenAI's Realtime API runs speech-to-speech models priced per token: gpt-realtime-2.1 at $32 per 1M audio tokens in ($0.40 cached) and $64 out, and gpt-realtime-2.1-mini at $10 in ($0.30 cached) and $20 out, plus text tokens. OpenAI also sells GPT-Live 1 voice sessions at $0.05 a minute, billed per second, with the backend model and tools charged on top. If you want a second option, the alternatives below are other speech-to-speech models (PolyAI's new Dialog-RSN-1 API, which speaks the same Realtime protocol, Google's Gemini 3.8 Live and Ultravox), platforms that run the whole call (Vapi, ElevenLabs Agents) and LiveKit's open-source agents framework. Prices are from each vendor's own pricing page.
Switching from OpenAI Realtime? PolyAI's Dialog-RSN-1 speaks the same protocol, so a first test can be as small as a new base URL and key. Run a few of your own calls before you compare minute prices.
PolyAI's Dialog-RSN-1 API, Gemini Live, Ultravox, Vapi, ElevenLabs Agents and LiveKit Agents compared with OpenAI's Realtime API on published pricing.
PolyAI- Starting price
- Platform: quote only. Dialog-RSN-1 API: 2¢/min plus text tokens
- Best for
- Enterprises automating high-volume customer service phone lines, or developers who want PolyAI's voice model on their own stack
Gemini Live API- Starting price
- Free tier; paid audio ~$0.005/min in, ~$0.018/min out (Gemini 3.8 Live)
- Best for
- Developers who want a realtime speech-to-speech model from Google
Ultravox- Starting price
- $0.05/min (first 30 min free)
- Best for
- Developers wanting a speech-native model at a flat per-minute rate
Vapi- Starting price
- $0 to start; $0.05/min Vapi hosting, models billed pass-through at cost
- Best for
- Developers who want to choose their own LLM, transcriber, voice and telephony provider
ElevenLabs Agents- Starting price
- Agents: $0.08 per minute
- Best for
- Builders who want ElevenLabs' voices and speech models in one agent pipeline
LiveKit Agents- Starting price
- Free; Ship from $50/mo
- Best for
- Developers building on open-source realtime infrastructure
What to check before you choose
- The quickest swap is PolyAI's Dialog-RSN-1 API. PolyAI's docs say it is wire-compatible with OpenAI's Realtime protocol: the same JSON events and function calling, with a new base URL and API key. It costs 2¢ per connected minute, billed per second with no minimums, plus text tokens at $2.00 in / $0.20 cached / $5.00 out per 1M. Voice output is free until Nov 19, and the page doesn't yet say what it costs after that. A session returns text (for your own voice) or audio, not both, and telephony and orchestration aren't included.
- Put prices on the same footing. OpenAI publishes no per-minute figure for gpt-realtime-2.1, so a minute's cost depends on how much each side talks. Google publishes per-minute equivalents for Gemini 3.8 Live on its paid tier: about $0.005 a minute of audio in and $0.018 a minute out, plus text at $0.75 in / $4.50 out per 1M tokens. Ultravox charges a flat $0.05 a minute that includes its model, after 30 free minutes (it says it has been acquired by Inworld; prices are unchanged).
- Decide whether you want a model or a whole call. OpenAI, PolyAI's API and Gemini Live give you the model; phone lines and call handling are up to you. Ultravox adds SIP at 0.5¢ a minute (0.48¢ on its $100/mo Pro plan). Vapi charges a $0.05/min hosting fee and bills transcriber, LLM and voice at cost; its own calculator estimates about $82–$129 for 1,000 minutes with Deepgram, OpenAI and ElevenLabs. ElevenLabs Agents is $0.08 a minute plus the LLM.
- Check concurrency before launch day. PolyAI's API allows 2 concurrent calls until you add a card, then 10 (more through sales). Ultravox's pay as you go stops at 5 (Pro removes the hard cap); Vapi starts at 4; LiveKit Cloud's free Build plan allows 5 concurrent agent sessions and Ship (from $50/mo) allows 20.
- Keep real callers off free tiers that train on your data. Google marks Gemini's free tier "used to improve our products"; its paid tier isn't.
- Want to own the pipeline? LiveKit Agents is open source. On LiveKit Cloud, Build is free with 1,000 agent session minutes and Ship starts at $50/mo with 5,000, then $0.01 a minute.
- Try before you switch: PolyAI gives $100 of API credit to sign-ups before Nov 19 (it was $20), Ultravox's first 30 minutes are free, Vapi gives $5 in credits, ElevenLabs' free plan includes 15 agent minutes, and LiveKit's Build plan includes $2.50 of inference credits (about 50 minutes, LiveKit says).
Not sure which fits? Answer a few quick questions in the finder →
Prices from vendors' official pricing pages, checked October 20, 2026. Not sponsored; order isn't a ranking.