← Back to community benchmarks
Kimi-Linear-48B-A3B-Instruct
Performance
4k
tokens
2,520
PP tok/s
91.9
TG tok/s
1626
TTFT (ms)
50.0
Peak mem (GB)
Hardware
Chip
M3 Ultra (80c)
Memory
96 GB
GPU Cores
80
Software
oMLX
v0.2.10
macOS
macOS 26.3.1
Context
4,096
Performance by Context Length
| Context | PP tok/s | TG tok/s | Peak Mem | |
|---|---|---|---|---|
| 1k | 2,207 | 92.5 | 49.7 GB | view |
| 4k | 2,520 | 91.9 | 50.0 GB | current |
| 8k | 2,532 | 89.6 | 50.3 GB | view |
| 16k | 2,320 | 86.3 | 51.0 GB | view |
| 32k | 1,926 | 71.9 | 52.4 GB | view |
| 64k | 1,408 | 58.2 | 55.2 GB | view |
Batching Results
| Batch Size | TG tok/s | Speedup |
|---|---|---|
| 1× | 92.5 | 1.00× |
| 2× | 153.3 | 1.66× |
| 4× | 223.2 | 2.41× |
| 8× | 326.7 | 3.53× |