← Back to community benchmarks
Qwen3.5-4B-Claude-4.6-Opus-Reasoning-Distilled-v2
Performance
32k
tokens
2,066
PP tok/s
72.4
TG tok/s
15860
TTFT (ms)
8.6
Peak mem (GB)
Hardware
Chip
M3 Ultra (80c)
Memory
512 GB
GPU Cores
80
Software
oMLX
v0.3.5
macOS
macOS 26.4.1
Context
32,768
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 1,573 | 105.1 | 5.3 GB | view |
| 4k | 2,178 | 101.7 | 6.0 GB | view |
| 8k | 2,364 | 96.6 | 6.3 GB | view |
| 16k | 2,277 | 86.7 | 7.1 GB | view |
| 32k | 2,066 | 72.4 | 8.6 GB | current |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 105.1 | 1.00× |
| 2× | 177.6 | 1.69× |
| 4× | 258.1 | 2.46× |