Benchmarks · community-measured
qwen3.8-27b-mixed-3.5bpw on Apple M5 Max
Real measured local-inference performance for
qwen3.8-27b-mixed-3.5bpw on an Apple M5 Max (36 GB unified memory),
from the rapid-mlx community benchmark corpus. Latest submission 2026-08-21,
measured on rapid-mlx 0.12.18.
26.5
tok/s decode
604
tok/s prefill
3.5 s
first token
16.7 GB
peak memory
A single community run so far — treat it as one data point, not a distribution. Reproduce it below and submit yours to firm it up.
All measurements for this pair
| RAM | Accel | Decode | Prefill | TTFT | Peak mem | Runs | rapid-mlx | Latest |
|---|---|---|---|---|---|---|---|---|
| 36 GB | — | 26.5 tok/s | 604 tok/s | 3.5 s | 16.7 GB | 1 | 0.12.18 | 2026-08-21 |
Standardized single-request benchmark: 5 measured rounds after 1 discarded warmup, greedy sampling, prefix cache disabled. Raw corpus: /api/benchmarks.
Reproduce this number
Install rapid-mlx, run the same suite on your Mac, and submit the result — it lands on the community leaderboard and updates this page.
$ curl -fsSL https://rapidmlx.com/install.sh | bash
$ rapid-mlx bench qwen3.8-27b-mixed-3.5bpw --tier speed --submit