M5 Pro LLM benchmark
No signed M5 Pro submissions yet. Check the configuration guidance below, or contribute a reproducible measurement from your own machine.
Specs vendor specs, not measured
- Unified memory (max configuration)
- 64 GB
Plan memory before choosing a model
The catalog lists a family maximum of 64 GB of unified memory. Enter your installed amount; smaller configurations have less space. Model weights, context cache and runtime overhead must share the available memory. Capacity alone does not establish which model will fit or how fast it will run. Use the memory-fit calculator for your model and context as an estimate, then verify with the intended runtime. No signed M5 Pro measurement is available yet.
Contribute a signed M5 Pro benchmark
Run on YOUR hardware to populate this page: pipx install llm-speed && llm-speed bench
$ pipx install llm-speed && llm-speed bench
Community folklore on M5 Pro
1 unverified claim extracted from Reddit/HN comments. Lower trust than signed runs above; every row links to the source.
- communityconfidence 50%
25.00tok/s — on M5 Pro via mlx
our signed data: M5 Pro
“bit benchmarks on M5 Pro 64GB (SwiftLM/MLX, measured):** |Config|Prefill|Decode|GPU RAM active| |:-|:-|:-|:-| |SSD streaming, 4K context|**25 t/s**|\~0.4 t/s|**2.7 GB**| Note: At 4,262-token context depth with a 122B MoE, each decode step streams the full active expert set (\~1…”
Request a run on M5 Pro
No measured run yet for these pairs. Vote for the one you want; the most-requested runs go first. See all requests.
We never sell or share your email. Privacy