What Block Says About M4 Max Support
The official page says inference is faster than real time and runs locally on a MacBook Pro M4 Max or better. These are developer-reported claims, not measurements from our own hardware tests.
What is not specified
The page we checked does not give minimum unified memory, storage size, a macOS version, supported quantization formats, or installation commands. We cannot infer those requirements from the chip name.
M1, M2, M3, Windows, and Linux
The published statement does not establish support for these systems. It also does not prove that other configurations are impossible. Wait for release documentation before buying hardware specifically for Dessa.
Training hardware is different
Block says training used one NVIDIA GB200 NVL72 system. That is a rack-scale training system, not the same thing as the Mac used for inference. Training hardware should not be read as a consumer requirement.
Before choosing hardware
Treat the published chip statement as a starting point, not a complete shopping list. A reproducible compatibility check needs the released model version, memory configuration, software version, audio setup and sustained conversation timing. Without these, a claim that another Mac is supported would be speculation.
Does local mean private?
Local inference can reduce dependence on a remote inference service, but privacy depends on the released software, telemetry, microphone permissions, and retention behavior. Those details require actual release documentation.