← Back to community benchmarks
Qwen3.5-35B-A3B-Claude-4.6-Opus-Reasoning-Distilled
Performance
16k
tokens
2,289
PP tok/s
65.5
TG tok/s
7158
TTFT (ms)
36.1
Peak mem (GB)
Hardware
Chip
M3 Ultra (80c)
Memory
512 GB
GPU Cores
80
Software
oMLX
v0.2.10
macOS
macOS 15.3
Context
16,384
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 1,658 | 77.5 | 35.3 GB | view |
| 4k | 2,337 | 74.9 | 35.5 GB | view |
| 8k | 2,406 | 71.0 | 35.7 GB | view |
| 16k | 2,289 | 65.5 | 36.1 GB | current |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 77.5 | 1.00× |
| 2× | 146.6 | 1.89× |
| 4× | 234.5 | 3.03× |
| 8× | 356.7 | 4.60× |