Compare
Loading community benchmarks…
Loading community benchmarks…
Put two chips, runtimes, or runtime versions side by side. Everything else is held fixed or called out, so a ratio is worth exactly what the matched runs say.
Ratios read side A relative to side B. Geometric mean over 2 like-for-like pairs.
Everything else matches: model, quantisation, backend, protocol, and workload are the same on both sides.
Each row is a configuration present on both sides. Values are per-cell medians; ratios read Apple M1 Max relative to Apple M3 Ultra.
| Configuration | Decode A | Decode B | Ratio | Prefill A | Prefill B | Ratio | Runs A / B |
|---|---|---|---|---|---|---|---|
Qwen3-4B-Thinking-2507 BaseRTQ4 | 93.2 | 155.6 | 0.60× | 921 | 2,582 | 0.36× | 1 / 1 |
Qwen3-4B-Thinking-2507 BaseRTQ8 | 71.5 | 118.2 | 0.60× | 925 | 2,559 | 0.36× | 1 / 1 |
Top 2 of 2 runs by decode throughput.
| Model / format | Chip / backend | Decode | Prefill | Contributor | Date | Report |
|---|---|---|---|---|---|---|
Qwen3-4B-Thinking-2507-Q4 BaseRT 0.2.6Q4 | Apple M1 Max Metal | 93.2 TG128 @ 1 ctx | 921 PP512 | basecompute | View Benchmark | |
Qwen3-4B-Thinking-2507-Q8 BaseRT 0.2.6Q8 | Apple M1 Max Metal | 71.5 TG128 @ 1 ctx | 925 PP512 | basecompute | View Benchmark |
Top 2 of 2 runs by decode throughput.
| Model / format | Chip / backend | Decode | Prefill | Contributor | Date | Report |
|---|---|---|---|---|---|---|
Qwen3-4B-Thinking-2507-Q4 BaseRT 0.2.6Q4 | Apple M3 Ultra Metal | 155.6 TG128 @ 1 ctx | 2,582 PP512 | basecompute | View Benchmark | |
Qwen3-4B-Thinking-2507-Q8 BaseRT 0.2.6Q8 | Apple M3 Ultra Metal | 118.2 TG128 @ 1 ctx | 2,559 PP512 | basecompute | View Benchmark |