← Back to community benchmarks
Qwen3-Coder-Next-oQ8e
Performance
32k
tokens
778.7
PP tok/s
54.0
TG tok/s
42078
TTFT (ms)
82.2
Peak mem (GB)
Hardware
Chip
M4 Max (40c)
Memory
128 GB
GPU Cores
40
Software
oMLX
v0.5.1
macOS
macOS 26.5.2
Context
32,768
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 1,231 | 66.0 | 80.2 GB | view |
| 4k | 1,327 | 64.6 | 80.9 GB | view |
| 8k | 1,182 | 62.8 | 81.2 GB | view |
| 16k | 993.9 | 58.9 | 81.6 GB | view |
| 32k | 778.7 | 54.0 | 82.2 GB | current |
| 64k | 481.5 | 30.5 | 82.7 GB | view |
| 128k | 326.4 | 28.5 | 84.3 GB | view |
| 195k | 240.7 | 21.0 | 85.8 GB | view |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 66.0 | 1.00× |
| 2× | 73.5 | 1.11× |
| 4× | 100.3 | 1.52× |
| 8× | 133.9 | 2.03× |