← Back to community benchmarks
Qwen3.5-27B
Performance
1k
tokens
415.9
PP tok/s
21.5
TG tok/s
2462
TTFT (ms)
29.2
Peak mem (GB)
Hardware
Chip
M3 Ultra (80c)
Memory
512 GB
GPU Cores
80
Software
oMLX
v0.2.19.dev2
macOS
macOS 26.3.1
Context
1,024
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 415.9 | 21.5 | 29.2 GB | current |
| 4k | 426.2 | 20.5 | 30.8 GB | view |
| 8k | 421.1 | 20.0 | 31.4 GB | view |
| 16k | 408.9 | 19.2 | 32.7 GB | view |
| 32k | 383.7 | 17.4 | 35.2 GB | view |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 21.5 | 1.00× |
| 2× | 38.3 | 1.78× |
| 4× | 56.6 | 2.63× |
| 8× | 69.3 | 3.22× |