← Back to community benchmarks
Qwen3-Coder-Next
Performance
8k
tokens
615.3
PP tok/s
39.8
TG tok/s
13313
TTFT (ms)
43.4
Peak mem (GB)
Hardware
Chip
M1 Ultra (48c)
Memory
64 GB
GPU Cores
48
Software
oMLX
v0.2.11
macOS
macOS 26.4
Context
8,192
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 440.0 | 45.0 | 43.0 GB | view |
| 4k | 601.7 | 43.4 | 43.2 GB | view |
| 8k | 615.3 | 39.8 | 43.4 GB | current |
| 16k | 599.4 | 36.0 | 43.9 GB | view |
| 32k | 552.1 | 29.3 | 44.7 GB | view |
| 64k | 451.8 | 19.9 | 46.5 GB | view |
| 128k | 311.0 | 6.2 | 50.1 GB | view |
| 195k | 223.9 | 2.5 | 54.0 GB | view |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 45.0 | 1.00× |
| 2× | 71.7 | 1.59× |
| 4× | 103.5 | 2.30× |
| 8× | 134.5 | 2.99× |