M3 Max Qwen 3 32B Q8_0 decode (Ollama, siliconscore ref)
7.53 tok/s (single-request decode, per-request)
Fields
id
benchmarks/m3max-qwen-3-32b-q8-0-ollama-ctx20172-tgMetric
decode_tok_s
Value
7.53
Unit
tok/s (single-request decode, per-request)
Model
Qwen 3 32B
Methodology
Silicon Score benchmarks.csv reference run. M3 Max (GPU count not published, 64 GB): Qwen 3 32B Q8_0 via Ollama. context_tokens=20172. Decode tok/s single-request.
Measured by
third_party
Reproducible
yes
Thermal state
unknown
notes
siliconscore reference run; source_url=https://www.reddit.com/r/LocalLLaMA/comments/1kk0ghi/speed_comparison_with_qwen332bq8_0_ollama/. Chip variant 'M3 Max (GPU count not published, 64 GB)'.
Notes
siliconscore reference run; source_url=https://www.reddit.com/r/LocalLLaMA/comments/1kk0ghi/speed_comparison_with_qwen332bq8_0_ollama/. Chip variant 'M3 Max (GPU count not published, 64 GB)'.