← Back to community benchmarks
Qwen3-Coder-Next
Performance
32k
tokens
1,549
PP tok/s
42.5
TG tok/s
21160
TTFT (ms)
81.8
Peak mem (GB)
Hardware
Chip
M3 Ultra (80c)
Memory
512 GB
GPU Cores
80
Software
oMLX
v0.2.6
macOS
macOS 15.5
Context
32,768
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 1,442 | 58.8 | 80.1 GB | view |
| 4k | 1,872 | 58.5 | 80.3 GB | view |
| 8k | 1,888 | 55.2 | 80.5 GB | view |
| 16k | 1,783 | 49.9 | 80.9 GB | view |
| 32k | 1,549 | 42.5 | 81.8 GB | current |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 58.8 | 1.00× |
| 2× | 103.3 | 1.76× |
| 4× | 166.3 | 2.83× |
| 8× | 260.7 | 4.43× |