OpenAI Realtime API vs Pipecat PhoneLLM

OpenAI Realtime API

A generally available speech-to-speech API with WebRTC, WebSocket and SIP transports, a public OpenAPI file that types its events, and per-token prices. Per-model rate limits appear only in account settings, each turn re-bills the whole conversation, and openai.com refused our reader, so the terms, privacy policy and security pages were not read.

Pipecat PhoneLLM

PhoneLLM Mini is callable per token on Workers AI at $0.06 input and $0.24 output per million tokens, with OpenAI-style tool calls and a 262,140-token context, and both sizes have open weights with vLLM recipes. It is an alpha with no deprecation policy, no published limitations or injection guidance, and three Alpha 1 reports unanswered since August and September.

OpenAI Realtime API

1. Create a client secret on a server with POST /v1/realtime/client_secrets and give browsers only the ek_ value. Never ship the API key.
2. After a response with MCP calls finishes, send another response.create. The API does not create the follow-up response itself.
3. Set truncation with a retention_ratio below 1 and token_limits.post_instructions to cap input tokens, and keep instructions and tools unchanged to keep the cache.
4. On a WebSocket, stop playback on input_audio_buffer.speech_started and send conversation.item.truncate with audio_end_ms. WebRTC and SIP truncate on the server.
5. Use gpt-realtime-2.1 or gpt-realtime-2.1-mini. gpt-realtime and gpt-realtime-mini shut down on 20 January 2027.

Pipecat PhoneLLM

1. Send temperature: 0 on every call. The repositories' generation_config.json samples by default, and Max's sets temperature 1.0
2. For Max, serve with --reasoning-parser nemotron_v3 and send chat_template_kwargs with enable_thinking: true on every request, or the reply lands in reasoning
3. For Mini, serve without a reasoning parser and keep enable_thinking false, as Cloudflare and the card say, or tool calls can be filed under reasoning
4. Confirm bookings, payments and transfers in your own handler before acting on a tool call, since caller speech reaches the model directly
5. Pin the model ID including alpha-2. Daily states no notice period, and a later alpha may replace it