← Back to community benchmarks
gpt-oss-120b-heretic-v2-mxfp4-q8-hi
Performance
16k
tokens
1,283
PP tok/s
49.9
TG tok/s
12767
TTFT (ms)
60.1
Peak mem (GB)
Hardware
Chip
M3 Ultra (80c)
Memory
512 GB
GPU Cores
80
Software
oMLX
v0.3.5
macOS
macOS 26.4.1
Context
16,384
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 848.6 | 66.0 | 59.5 GB | view |
| 4k | 1,294 | 68.4 | 59.6 GB | view |
| 8k | 1,336 | 59.8 | 59.7 GB | view |
| 16k | 1,283 | 49.9 | 60.1 GB | current |
| 32k | 1,141 | 45.1 | 60.5 GB | view |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 66.0 | 1.00× |
| 2× | 87.6 | 1.33× |
| 4× | 113.1 | 1.71× |