← Back to community benchmarks
Qwen3-235B-A22B-Instruct-25076.5bit
Performance
16k
tokens
292.7
PP tok/s
14.9
TG tok/s
55982
TTFT (ms)
181.8
Peak mem (GB)
Hardware
Chip
M3 Ultra (80c)
Memory
512 GB
GPU Cores
80
Software
oMLX
v0.2.6
macOS
macOS 26.2
Context
16,384
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 360.5 | 22.5 | 178.8 GB | view |
| 4k | 376.0 | 20.1 | 179.6 GB | view |
| 8k | 346.8 | 17.5 | 180.4 GB | view |
| 16k | 292.7 | 14.9 | 181.8 GB | current |
| 32k | 221.1 | 11.1 | 184.7 GB | view |
| 64k | 146.5 | 7.3 | 190.7 GB | view |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 22.5 | 1.00× |
| 2× | 33.4 | 1.48× |
| 4× | 44.8 | 1.99× |