← Back to community benchmarks
Qwen3-Coder-Next-oQ2.7e
Performance
64k
tokens
537.5
PP tok/s
30.0
TG tok/s
121925
TTFT (ms)
33.9
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
65,536
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 125.9 | 17.9 | 70.0 GB | view |
| 4k | 1,088 | 67.4 | 66.9 GB | view |
| 8k | 1,110 | 67.3 | 66.0 GB | view |
| 16k | 991.4 | 62.2 | 32.7 GB | view |
| 32k | 838.7 | 59.9 | 33.3 GB | view |
| 64k | 537.5 | 30.0 | 33.9 GB | current |
| 128k | 290.7 | 28.3 | 35.1 GB | view |
| 195k | 219.4 | 23.2 | 36.9 GB | view |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 17.9 | 1.00× |
| 2× | 110.6 | 6.18× |
| 4× | 136.4 | 7.62× |
| 8× | 184.1 | 10.28× |