llms.txt
+14 −3
− USD 0.03 per connected AI minute, billed per second, plus AI provider and carrier usage. Minimum prepaid top-up USD 10. No monthly subscription, trial or free credits. Customer-owned provider usage is not charged twice. New workspace access is currently limited; contact mariano@apifleet.dev.− - [Build voice agents with API & MCP](https://apifleet.dev/index.md): Build, test and operate AI voice agents. Choose providers, connect tools and inspect calls. Pay $0.03 per AI minute plus provider and carrier usage.− - [Pay per second for voice agents](https://apifleet.dev/pricing.md): APIFleet costs $0.03 per connected AI minute plus provider and carrier usage. Prepaid top-ups start at $10, with no subscription or free credits.+ Platform fee: USD 0.03 per connected AI minute, billed per second (Vapi’s comparable hosting fee: USD 0.05, checked 2026-10-29). AI provider and carrier usage are billed on top. Minimum prepaid top-up USD 10. No monthly subscription, trial or free credits. Customer-owned provider usage is not charged twice. New workspace access is currently limited; contact mariano@apifleet.dev.+ - [Build voice agents with API & MCP](https://apifleet.dev/index.md): Build, test and operate AI voice agents. Choose providers, connect tools and inspect calls. $0.03/min platform fee vs Vapi’s $0.05, plus providers and carrier.+ - [Pay per second for voice agents](https://apifleet.dev/pricing.md): APIFleet’s platform fee is $0.03 per connected AI minute; Vapi’s is $0.05. AI provider and carrier usage come on top of both. Top-ups from $10.+ - [gpt-live-1 on Real Phone Calls: A Field Guide](https://apifleet.dev/blog/gpt-live-1.md): gpt-live-1 measured on 1,073 APIFleet phone calls: greeting time, backend and tool timing, error-ended calls and cost per connected minute, with sample sizes.+ - [Piper TTS on phone calls: latency and cost guide](https://apifleet.dev/blog/piper-tts.md): Self-hosted Piper TTS on 271 APIFleet phone calls: reply latency, first audio, greeting time, errors and cost per minute, with sample sizes and limits.+ - [Speech-to-Speech Models in Production: A Guide](https://apifleet.dev/blog/speech-to-speech-models.md): Speech-to-speech models on 1,908 APIFleet calls: response latency, openings, errors, cost and tokens per minute for gpt-live-1, Gemini Live and gpt-realtime.+ - [TTS Models for Phone Calls: Production Data Guide](https://apifleet.dev/blog/tts-models-phone-calls.md): TTS models timed on APIFleet phone calls: first audio, reply latency, errors and cost per minute for Piper, Deepgram and Cartesia, plus Soniox in the browser.+ - [Voice Agent Evaluation: How to Run an A/A Test](https://apifleet.dev/blog/voice-agent-evaluation-noise.md): Voice agent evaluation needs a noise floor: identical configs scored 2.91 points apart on a partly LLM-judged scale. See APIFleet's A/A data and sample sizes.+ ## Reddit answers+ - [Reddit voice-agent questions, answered with data](https://apifleet.dev/reddit.md): Voice-agent questions from Reddit, answered with aggregate data from calls on APIFleet. Independent answers, not affiliated with or endorsed by Reddit.+ - [Barge-in During Call Opening: Who Speaks First?](https://apifleet.dev/reddit/barge-in-during-call-opening.md): Barge-in during call opening: how often callers spoke before an AI agent's first audio on APIFleet phone calls, inbound vs outbound, and consent-line options.+ - [Fastest TTS for Phone Calls? Production Timings](https://apifleet.dev/reddit/fastest-tts-for-phone-calls.md): Fastest TTS for phone calls, timed on APIFleet production calls: first audio, reply latency, errors and cost per minute for Piper, Deepgram and Cartesia.+ - [Gemini Live vs OpenAI Realtime: phone latency](https://apifleet.dev/reddit/gemini-live-vs-openai-realtime.md): Gemini Live vs OpenAI Realtime on production phone calls: reply latency per model, the 5 September Gemini stalls, error reporting gaps and cost per minute.+ - [Voice agent tool call latency: tool or model?](https://apifleet.dev/reddit/voice-agent-tool-call-latency.md): Voice agent tool call latency on APIFleet production calls: how long the tool, the model passes and voice start took on tool turns, with samples and dates.