Speko
One API for voice AI models
Speko routes voice agents to the best speech-to-text, language and text-to-speech models for each language and use case through one API, with public benchmarks.
- Founder
- Beknazar Abdikamalov
- YC batch
- S26
Company posts
Summaries from the company’s own website.
- Gemini 3.8 Flash Is the Most Robust TTS Model on Our Board ↗
Speko Benchmarks says Google’s Gemini 3.8 Flash and Flash-Lite TTS models both rank in the top naturalness group on its English board. The post reports that Flash handled numbers best in their tests, while Flash-Lite was faster but made more mistakes with minus signs.
- Gemini 3.8 Live Tops Our Board. Its Smarter Sibling Comes Fourth. ↗
The post says Google released Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking, and that Speko benchmarked both on a six-scenario concierge harness. It reports that Gemini 3.8 Live scored highest on the speech-to-speech board with faster answers, while the Extended Thinking version reasoned better in some respects but answered much more slowly across reasoning levels.