RXed AI News

AI news with a spine.
@RXed_EU
SCORE 6.4/1018/20 elements Audited 2026-07-25 · RXed table v1.0

Vapi

Real infrastructure behind the $50M raise and the Amazon Ring deployment — but budget 3-7x the advertised $0.05/minute once STT, LLM, TTS and telephony all send separate invoices.

SpecialistAutomation & AgentsVoice & SpeechSalesPaid
Vendor
Vapi, Inc. · vapi.ai
Origin
US — San Francisco
Pricing
Build (pay-as-you-go): ~€0.044/min (~$0.05) hosting + model-provider costs at cost (€0 with your own API key); 10 concurrent calls included, +~€8.75/line/mo (~$10); HIPAA add-on ~€1,750/mo (~$2,000); Zero Data Retention add-on ~€875/mo (~$1,000) · Scale: annual contract, custom volume-based pricing (SOC2, HIPAA, PCI, SSO, RBAC, dedicated support). Verified on vapi.ai/pricing 25/07/2026; USD-denominated, converted at 1 USD = 0.875 EUR.
Users (official only)
1M+ developers, 2.7M+ unique agents created, 1B+ calls processed (source, 2026-05-12)

Element scores

Reactive
Retrieval & Memory
Orchestration
Validation
Models
Primitives
Pr7.5
Prompts
Em4
Embeddings
Cx6.5
Context
Tr6.5
Tracing
Lg8
LLM
Compositions
Fc8.5
Function calling
Vx3
Vector store
Rg6
RAG
Gr6
Guardrails
Mm5
Multimodal
Deployment
Ag8.5
Agents
Ft4
Fine-tuning
Fw8
Frameworks & harnesses
Ev7
Evaluations
Sm6
Small models
Emerging
Ma8
Multi-agent
Sy
Synthetic data
Pc7.5
Protocols
In
Interpretability
Th5
Thinking models
Tap or hover any element to see why it got that score.

Strengths

Broad bring-your-own-model support across every major LLM provider, a mature function-calling/MCP surface, and Squads for production multi-agent handoffs (240,000 calls/day at one customer). Flow Studio, Workflows, official SDKs and a CLI form a real harness, and native Evals, Simulations and Monitoring (GA April 2026) give teams in-platform testing without a third-party eval layer. Amazon Ring routing 100% of inbound support calls through Vapi after evaluating 40+ vendors is a serious production proof point, backed by a $50M Series B at a ~$500M valuation.

Honest dings

The advertised $0.05/minute is the platform fee only — independent breakdowns (CloudTalk, Retell, Zeeg, Rezora) consistently land the true all-in cost at $0.13-0.36/minute once STT, LLM, TTS and telephony are counted, and HIPAA plus Zero Data Retention add $3,000/month before any usage. Trustpilot sits at 2.4/5 with recurring billing-surprise and slow-support complaints. No native vector store or embeddings — RAG leans entirely on a third-party index (LlamaCloud).

Best for: Engineering teams that want full programmatic control over a production voice-agent stack and are willing to own a multi-vendor billing and ops surface to get it.

Sources

Every audit lists the research it rests on — transparency and traceability are the product. Tools evolve: each audit is a snapshot of its audit date, and re-audits supersede older versions (kept below for reference).