← Back to community benchmarks
Qwen3.5-4B
Performance
8k
tokens
376.8
PP tok/s
30.6
TG tok/s
21742
TTFT (ms)
4.6
Peak mem (GB)
Hardware
Chip
M4 (10c)
Memory
16 GB
GPU Cores
10
Software
oMLX
v0.2.6
macOS
macOS 26.3
Context
8,192
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 406.0 | 38.9 | 3.9 GB | view |
| 4k | 397.4 | 36.4 | 4.2 GB | view |
| 8k | 376.8 | 30.6 | 4.6 GB | current |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 38.9 | 1.00× |
| 2× | 74.5 | 1.92× |
| 4× | 93.6 | 2.41× |
| 8× | 95.8 | 2.46× |