llamaperf
Calculator
GPUs
Engines
Guides
Blog
Community
MCP
Advertise
Submit
Sign in
Submit
llamaperf
/
M4 Max
/
report
Bonsai 2 2B
on M4 Max
Oct 16, 2026
Share
X
Reddit
LinkedIn
Bluesky
Copy
link
Throughput
500.0 t/s gen
Quant
ternary
Summary
User reports the release of Ternary Bonsai 2 (2B, 4B, and 9B) on Hugging Face. The 2B model runs entirely locally in the browser at up to 500 tok/s on an M4 Max using custom WebGPU kernels.
View source ↗
Feedback