← Back to community benchmarks
Qwen3-235B-A22B
Performance
8k
tokens
347.4
PP tok/s
15.1
TG tok/s
23583
TTFT (ms)
235.1
Peak mem (GB)
Hardware
Chip
M3 Ultra (80c)
Memory
512 GB
GPU Cores
80
Software
oMLX
v0.2.7
macOS
macOS 15.7.4
Context
8,192
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 363.0 | 19.2 | 233.5 GB | view |
| 4k | 379.9 | 17.0 | 234.4 GB | view |
| 8k | 347.4 | 15.1 | 235.1 GB | current |
| 16k | 292.9 | 12.9 | 236.5 GB | view |
| 32k | 219.2 | 10.1 | 239.4 GB | view |
| 64k | 143.7 | 6.6 | 245.4 GB | view |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 19.2 | 1.00× |
| 2× | 27.2 | 1.42× |
| 4× | 36.0 | 1.88× |
| 8× | 65.3 | 3.40× |