← Back to community benchmarks
Qwen3.5-27B
Performance
32k
tokens
383.4
PP tok/s
25.4
TG tok/s
85457
TTFT (ms)
20.1
Peak mem (GB)
Hardware
Chip
M3 Ultra (80c)
Memory
512 GB
GPU Cores
80
Software
oMLX
v0.2.6
macOS
macOS 26.3.1
Context
32,768
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 423.4 | 35.7 | 16.6 GB | view |
| 4k | 429.2 | 34.0 | 17.0 GB | view |
| 8k | 423.5 | 32.5 | 17.4 GB | view |
| 16k | 409.8 | 29.7 | 18.3 GB | view |
| 32k | 383.4 | 25.4 | 20.1 GB | current |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 35.7 | 1.00× |
| 2× | 65.1 | 1.82× |
| 4× | 89.3 | 2.50× |
| 8× | 102.9 | 2.88× |