← Back to community benchmarks
gemma-4-31B-it
Performance
64k
tokens
94.7
PP tok/s
4.9
TG tok/s
691922
TTFT (ms)
45.1
Peak mem (GB)
Hardware
Chip
M3 Max (40c)
Memory
128 GB
GPU Cores
40
Software
oMLX
v0.3.8
macOS
macOS 15.7.4
Context
65,536
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 147.5 | 9.9 | 31.8 GB | view |
| 4k | 161.5 | 10.1 | 34.5 GB | view |
| 8k | 154.2 | 8.4 | 35.0 GB | view |
| 16k | 116.8 | 5.7 | 36.1 GB | view |
| 32k | 108.4 | 5.3 | 39.1 GB | view |
| 64k | 94.7 | 4.9 | 45.1 GB | current |
| 128k | 73.8 | 3.6 | 57.1 GB | view |
| 195k | 56.8 | 2.7 | 61.4 GB | view |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 9.9 | 1.00× |
| 2× | 11.6 | 1.17× |
| 4× | 15.6 | 1.58× |
| 8× | 18.0 | 1.82× |