← Back to community benchmarks
Qwen3.5-4B
Performance
64k
tokens
1,056
PP tok/s
45.5
TG tok/s
62090
TTFT (ms)
9.9
Peak mem (GB)
Hardware
Chip
M2 Ultra (76c)
Memory
128 GB
GPU Cores
76
Software
oMLX
v0.2.6
macOS
macOS 26.3
Context
65,536
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 1,423 | 99.9 | 5.9 GB | view |
| 4k | 1,448 | 96.2 | 6.1 GB | view |
| 8k | 1,421 | 90.1 | 6.5 GB | view |
| 16k | 1,362 | 80.1 | 7.3 GB | view |
| 32k | 1,255 | 63.0 | 7.9 GB | view |
| 64k | 1,056 | 45.5 | 9.9 GB | current |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 99.9 | 1.00× |
| 2× | 173.4 | 1.74× |
| 4× | 235.7 | 2.36× |