← Back to community benchmarks
Qwen3.5-27B-Claude-4.6-Opus-Distilled
Performance
64k
tokens
150.5
PP tok/s
11.8
TG tok/s
435422
TTFT (ms)
26.7
Peak mem (GB)
Hardware
Chip
M2 Ultra (60c)
Memory
64 GB
GPU Cores
60
Software
oMLX
v0.3.5.dev1
macOS
macOS 26.3
Context
65,536
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 175.0 | 33.1 | 15.8 GB | view |
| 4k | 182.1 | 31.0 | 17.3 GB | view |
| 8k | 181.5 | 29.1 | 17.9 GB | view |
| 16k | 178.4 | 26.5 | 19.1 GB | view |
| 32k | 170.6 | 21.0 | 21.6 GB | view |
| 64k | 150.5 | 11.8 | 26.7 GB | current |
| 128k | 121.8 | 6.2 | 37.0 GB | view |
| 195k | 100.2 | 4.6 | 47.8 GB | view |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 33.1 | 1.00× |
| 2× | 40.9 | 1.24× |
| 4× | 45.6 | 1.38× |
| 8× | 48.8 | 1.47× |