← Back to community benchmarks
NVIDIA-Nemotron-3-Nano-30B-A3B
Performance
32k
tokens
1,012
PP tok/s
81.8
TG tok/s
32377
TTFT (ms)
21.7
Peak mem (GB)
Hardware
Chip
M2 Ultra (60c)
Memory
128 GB
GPU Cores
60
Software
oMLX
v0.5.1
macOS
macOS 26.5
Context
32,768
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 992.7 | 95.3 | 20.3 GB | view |
| 4k | 1,137 | 93.4 | 21.3 GB | view |
| 8k | 1,137 | 91.3 | 21.5 GB | view |
| 16k | 1,093 | 87.3 | 21.6 GB | view |
| 32k | 1,012 | 81.8 | 21.7 GB | current |
| 64k | 878.8 | 71.8 | 22.2 GB | view |
| 128k | 690.2 | 58.9 | 22.7 GB | view |
| 195k | 606.2 | 49.4 | 23.3 GB | view |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 95.3 | 1.00× |
| 2× | 113.6 | 1.19× |
| 4× | 178.4 | 1.87× |
| 8× | 246.0 | 2.58× |