← Back to community benchmarks
DeepSeek-R1
Performance
4k
tokens
217.1
PP tok/s
18.1
TG tok/s
18897
TTFT (ms)
395.9
Peak mem (GB)
Hardware
Chip
M3 Ultra (80c)
Memory
512 GB
GPU Cores
80
Software
oMLX
v0.2.11
macOS
macOS 26.3.1
Context
4,096
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 215.2 | 19.0 | 392.0 GB | view |
| 4k | 217.1 | 18.1 | 395.9 GB | current |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 19.0 | 1.00× |
| 2× | 25.2 | 1.33× |
| 4× | 31.0 | 1.63× |