Official provider results
Dessa launch recordings
Block shares live recordings of people talking with Dessa, a full-duplex model trained only on human speech. Block publishes no benchmark scores; the model is waitlist-only and coming soon.
Source snapshots refreshed
Five measurement lanesVOICE / S2S
Reasoning, task completion, conversation dynamics, experience, and latency stay separate.
Benchmark profile
- Scope
- Unscripted live conversations recorded by Block, with subtitles added afterwards
- Primary metric
- None published (recordings and capability claims only)
- Owner
- Block
- Available evidence
- Launch page with 6 featured and 15 additional recordings, October 2026
What it measures
Conversation dynamics
Can it handle turns, interruptions, ambiguity, and state?
Voice experience
Is the exchange natural, robust, and responsive?
Available results
- Quantitative evaluation
- Not published
- Inference speed
- Faster than real time
- Local hardware
- MacBook Pro M4 Max or better
- Training compute
- One NVIDIA GB200 NVL72
Coming soon
Block publishes recordings, not benchmark tables.
Block claim
Stated on the launch page; no latency figure given.
Block claim
Stated on the launch page.
Block claim
Trained from scratch on human conversation audio with no text and no pretrained models.
Interpretation limit
Block chose and recorded these conversations, and it reports no quantitative evaluation. Faster-than-real-time inference, local MacBook Pro M4 Max inference, and single GB200 NVL72 training are Block claims that have not been independently checked. Nothing here enters weighted text-model scores.