← Back to community benchmarks
qwen-opus
Performance
32k
tokens
1,354
PP tok/s
68.7
TG tok/s
24205
TTFT (ms)
30.6
Peak mem (GB)
Hardware
Chip
M4 Max (40c)
Memory
128 GB
GPU Cores
40
Software
oMLX
v0.3.2
macOS
macOS 26.3.1
Context
32,768
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 1,323 | 95.3 | 27.3 GB | view |
| 4k | 1,656 | 92.1 | 28.3 GB | view |
| 8k | 1,646 | 84.6 | 28.6 GB | view |
| 16k | 1,549 | 77.3 | 29.3 GB | view |
| 32k | 1,354 | 68.7 | 30.6 GB | current |
| 64k | 1,048 | 50.8 | 33.2 GB | view |
| 128k | 688.6 | 34.9 | 38.6 GB | view |
| 195k | 502.7 | 23.9 | 44.3 GB | view |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 95.3 | 1.00× |
| 2× | 146.4 | 1.54× |
| 4× | 195.5 | 2.05× |