Kimi K2.6
moonshotai/Kimi-K2.6
Parameters
1000B
Context
256K
Downloads
785.6K
Likes
1.6K
Architecture
moe_transformer
GPU Requirements
VRAM / RAM / disk estimates by quantization and usage scenario
| Quantization | Minimum | Recommended | Production |
|---|---|---|---|
| FP1616bit | 2361.2 GB RAM 3541.8 · Disk 2944.7 | 2484.1 GB RAM 3726.1 · Disk 2970.7 | 2688.1 GB RAM 4032.1 · Disk 3009.7 |
| FP88bit | 1183.1 GB RAM 1774.6 · Disk 1491.9 | 1252.4 GB RAM 1878.6 · Disk 1517.9 | 1402.8 GB RAM 2104.2 · Disk 1556.9 |
| INT88bit | 1183.1 GB RAM 1774.6 · Disk 1491.9 | 1252.4 GB RAM 1878.6 · Disk 1517.9 | 1402.8 GB RAM 2104.2 · Disk 1556.9 |
| AWQ4bit | 616.1 GB RAM 924.2 · Disk 792.7 | 659.6 GB RAM 989.5 · Disk 818.7 | 784.3 GB RAM 1176.5 · Disk 857.7 |
| GPTQ4bit | 623.5 GB RAM 935.2 · Disk 801.8 | 667.3 GB RAM 1001.0 · Disk 827.8 | 792.3 GB RAM 1188.5 · Disk 866.8 |
| GGUF4bit | 638.2 GB RAM 957.3 · Disk 819.9 | 682.7 GB RAM 1024.1 · Disk 845.9 | 808.4 GB RAM 1212.6 · Disk 884.9 |
GPU Recommendations
Best ValueBEST
8× H200 SXM
1128 GB VRAM
$20,440/mo
≈ 31.8 tok/s
Effective production throughput ≈ 22.3 tok/s
score 0.3757
Cheapest
4× B200
720 GB VRAM
$12,051/mo
≈ 0 tok/s (N/A — benchmark unavailable)
Effective production throughput ≈ 0 tok/s
score 0.3057
Performance
8× H200 SXM
1128 GB VRAM
$20,440/mo
≈ 31.8 tok/s
Effective production throughput ≈ 22.3 tok/s
score 0.3757
Min Complexity
4× B200
720 GB VRAM
$12,051/mo
≈ 0 tok/s (N/A — benchmark unavailable)
Effective production throughput ≈ 0 tok/s
score 0.3057
💬 Community — real deployment experience
0 Articles · 0 Benchmarks