LLMCheck: Apple M5 Max (128GB), Ollama, Llama 3.3 70B Q4_K_M decode
12.0 tok/s (decode, per-request)
Fields
id
benchmarks/llmcheck-m5-max-llama33-70b-q4km-ollama-decodeMetric
decode_tok_s
Value
12
Unit
tok/s (decode, per-request)
Model
Llama 3.3 70B
Methodology
THIRD-PARTY MEASURED (LLMCheck open dataset, 2026-03..09 rows). Apple M5 Max 128GB unified memory, Ollama engine, Llama 3.3 70B Q4_K_M. avg tok/s for a chat-style generation; TTFT 2.8 s. Per-request decode, batch 1.
Measured by
third_party
Reproducible
yes
notes
Replaces the previous estimate-band figure (~15-25 tok/s) with a measured one. Matches the M5 Max bandwidth story (614 GB/s on a ~35 GB Q4 weight read -> low-teens decode).
Notes
Replaces the previous estimate-band figure (~15-25 tok/s) with a measured one. Matches the M5 Max bandwidth story (614 GB/s on a ~35 GB Q4 weight read -> low-teens decode).