Vapi alternatives for developers: buy the platform or own the stack

Vapi sells a managed phone platform. Some alternatives sell the same kind of platform. Others replace one layer of it and leave the rest to you.

Duplex Compute · October 13, 2026 · Rates and versions checked October 13, 2026

Before you compare vendors, decide which layer you want to stop paying for. A Vapi agent has a phone line, an orchestration loop that handles turns and interruptions, and models that hear, reason and speak. A managed platform runs all three. An agent framework gives you the loop as code. A voice model API gives you the hearing and speaking, and you build everything around it.

Teams with phone numbers, call transfers and staff who edit agents in a dashboard usually belong on a platform. Teams adding voice to their own web or mobile app often pay platform fees for phone features they never use.

What Vapi charges today

Vapi's usage-only plan charges $0.05 per minute for hosting, and passes model costs through at cost. Its calculator lists Deepgram transcription at $0.0095 to $0.0099 per minute, OpenAI models at $0.0077 to $0.0452, and ElevenLabs voices at $0.0146 to $0.0238. For 1,000 minutes it estimates $82 to $129 a month: $50 of hosting, $10 of transcription, $8 to $45 of LLM and $15 to $24 of voice.

The usage-only plan includes four concurrent calls, one Vapi phone number, $5 of starting credit and 14 days of raw data retention. Extra concurrent lines cost $10 per line per month. The Core package costs $29 a month and raises the allowance to 10 concurrent calls. Pro charges 10% of the hosting fee with a $999 monthly minimum and includes 30. A HIPAA add-on costs $2,000 a month.

Transport is billed by the carrier, not Vapi. The pricing page lists Vapi telephony, SIP, WebSockets and Daily WebRTC as free, Twilio inbound at $0.008 per minute and Twilio outbound at $0.014.

Older comparison articles quote different concurrency allowances. The figures above come from the live pricing page on October 13, 2026.

The alternatives, sorted by layer

The voice agent pricing hub converts these units into one per-minute figure and has a calculator for monthly totals.

Vapi vs Retell

Both are managed platforms for phone agents, and both bill per call minute. They split the bill differently. Vapi adds a flat hosting fee to models you pick from its catalog. Retell charges $0.055 per minute for voice infrastructure and lists its own per-minute price for each voice and each LLM.

With Retell Platform Voices at $0.015 and GPT-4.1 mini at $0.0128, Retell's AI charge is $0.0828 per minute. Vapi's cheapest calculator combination comes to $0.0818. The two configurations use different models, so the near-match says only that both start around eight cents a minute before telephony. Retell AI pricing works through three Retell configurations at 10,000 minutes.

Concurrency separates them more clearly. Retell includes 20 concurrent calls and charges $8 per month for each additional one. Vapi's usage plan includes four, at $10 per extra line. An app that peaks at 30 simultaneous calls pays Retell $80 a month for concurrency. On Vapi it pays $260 on the usage plan, or $229 on Core ($29 plus 20 extra lines).

Retell documents billing rules that matter for long or complex calls. Silence and hold time are billable, and the AI fee stops when a call transfers while telephony keeps running. Prompts over 4,000 tokens scale the billed duration, and that count includes the transcript and tool results.

Vapi vs LiveKit

This comparison is about ownership. LiveKit Agents is an open-source framework, Apache 2.0, with Python and Node.js SDKs. You write the agent in code, choose providers through plugins, and deploy to LiveKit Cloud or your own servers. On Vapi you configure an agent inside a hosted product.

The meters are smaller. LiveKit's Ship plan costs $50 a month, includes 5,000 agent-session minutes and 20 concurrent agent sessions, then charges $0.01 per session minute. LiveKit's own calculator shows $0.0479 per minute for a phone call: $0.01 agent session, $0.01 telephony, $0.0014 LLM, $0.0075 STT, $0.009 TTS and $0.01 observability. Vapi's hosting fee alone is $0.05.

You pay for the smaller meter with work. Your team owns the agent code, the deploys, the turn-detection settings and the debugging when a call goes wrong. LiveKit handles telephony through its own phone numbers or a third-party SIP trunk, so phone apps stay possible. Pipecat sits at the same layer as an open-source Python framework, and it can run over LiveKit transport. Pipecat vs. LiveKit compares the two frameworks, and LiveKit pricing builds complete bills.

Where a voice model API fits

GPT-Live, GPT-Realtime-2.1, Gemini 3.8 Live, Deepgram's Voice Agent API and Duplex replace the part of the stack that hears and speaks. None of them gives you a phone number, call transfers, IVR navigation or a dashboard for non-engineers. A web or mobile app with its own interface and backend may not need any of those.

Their billing units differ more than their prices suggest. GPT-Live bills $0.05 per minute of session, and silence and backend work count as active time. Realtime bills tokens, and the whole conversation is sent again on each response. Gemini re-bills retained context every turn, and proactive audio is permanently on for Gemini 3.8 Live, so listening time costs input tokens even when the model stays quiet.

Duplex Compute is a full-duplex voice API with tool calling, configurable behavior and long conversations, priced at $0.01 per connected minute. It belongs at this layer: inside your own app, or behind telephony you run. It does not replace Vapi's phone operations. The rate covers the voice session. Your tool handlers and any carrier minutes cost what they cost.

Which layer to pick

Whichever layer you choose, compare the cost of a completed task, not the cost of a minute. A cheap minute that loses a corrected booking still costs the call. Voice agent evaluation covers how to measure that.

Related notes

• Voice agent pricing: calculator and billing formulas
• Retell AI pricing
• Pipecat vs. LiveKit
• LiveKit pricing for voice agents
• OpenAI Realtime API alternatives