← Back to community benchmarks
Qwen3.5-397B-A17B
Performance
64k
tokens
356.8
PP tok/s
13.0
TG tok/s
183700
TTFT (ms)
308.5
Peak mem (GB)
Hardware
Chip
M3 Ultra (80c)
Memory
512 GB
GPU Cores
80
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 | 443.4 | 29.2 | 302.7 GB | view |
| 4k | 492.2 | 28.3 | 302.9 GB | view |
| 8k | 491.1 | 27.0 | 303.3 GB | view |
| 16k | 471.0 | 24.5 | 304.0 GB | view |
| 32k | 431.7 | 21.2 | 305.5 GB | view |
| 64k | 356.8 | 13.0 | 308.5 GB | current |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 29.2 | 1.00× |
| 2× | 45.7 | 1.57× |
| 4× | 57.9 | 1.98× |
| 8× | 71.9 | 2.46× |