Speko
Every voice model is benchmarked in English. Your users don't speak English. Speko measures 56 speech and language models across 10 languages.
Last verified:
What is Speko?
Speko is a provider-neutral voice AI router that benchmarks and routes speech-to-text, LLM, and text-to-speech models across multiple providers (OpenAI, Google, AssemblyAI, etc.). It offers a hosted API at router.speko.dev and an open-source Gateway runtime for frameworks like LiveKit and Pipecat, with native integrations for Claude and Cursor MCP.
Speko pricing
Pricing model: Freemium
Speko pros
- Multi-provider model routing with language-by-language benchmarking across 9+ languages
- Transparent cost vs. accuracy comparison table showing 20+ STT models
- Open-source Gateway runtime supporting LiveKit, Pipecat, and MCP protocols
- Bring-your-own-credentials (BYOK) option for provider-direct streaming without vendor lock-in
Speko cons
- No visible pricing information, free tier details, or subscription tiers listed
- Limited documentation on reliability, uptime SLAs, or rate limits in the provided content
- Requires API key signup; no public/anonymous access
Frequently asked questions about Speko
What providers does Speko support?
Speko integrates with OpenAI, Google, AssemblyAI, Alibaba, DeepGram, ElevenLabs, Cartesia, and 10+ other STT/TTS/LLM providers.
What languages are supported?
English, Arabic, French, German, Hindi, Norwegian, Spanish, Tamil, and Telugu across benchmarked models.
Can I use my own API keys?
Yes, Speko supports local bring-your-own-credentials (BYOK) for provider-direct streaming.
What frameworks integrate with Speko?
LiveKit, Pipecat, and Claude/Cursor MCP, with REST/OpenAPI and AsyncAPI contracts available.