Speko launches benchmark-driven voice AI router

Speko’s strongest wedge is operational: it turns voice-model selection from a one-time R&D project into a continuously measured routing decision. The approach is compelling for multilingual teams, though benchmark coverage and provider drift will determine its long-term value.

• –Public accuracy, latency, and cost scores make model selection more actionable than vendor marketing.
• –One API spans STT, LLM, and TTS providers, reducing integration lock-in.
• –The open-source gateway supports self-hosting and BYOK deployments with direct provider connections.
• –Failover happens during connection setup, so teams should assess resilience during active-call failures.
• –Pricing starts at 5% above provider rates or $0.09 per minute for bundled infrastructure.