Official provider results
Pocket TTS project notes
Kyutai’s TTS page and repository document Pocket TTS’s size, CPU real-time factor, and first-audio latency.
Source snapshots refreshed
Five measurement lanesVOICE / S2S
Reasoning, task completion, conversation dynamics, experience, and latency stay separate.
Benchmark profile
- Scope
- Parameter count, CPU throughput, and time to first audio chunk
- Primary metric
- Latency and real-time factor
- Owner
- Kyutai
- Available evidence
- Documented figures only; no listening-test result
What it measures
Latency
How long do responses, tool calls, and full tasks take?
Available results
- Parameters
- 100M
- CPU speed
- ~6x real time on MacBook Air M4
- Time to first audio
- ~200 ms
Documented
Runs on two CPU cores.
Documented
About 2.3–2.5x on a Tesla T4 CPU path, 6.28x on its GPU.
Documented
Streaming generation.
Interpretation limit
Provider-documented performance on reference hardware; no quality benchmark is published.