OCR / VisionMIT
Command A+
CohereLabs/command-a-plus-05-2026-bf16
Parameters
218B
Context
200K
Downloads
86.7K
Likes
141
Architecture
moe_transformer
GPU Requirements
VRAM / RAM / disk estimates by quantization and usage scenario
| Quantization | Minimum | Recommended | Production |
|---|---|---|---|
| FP1616bit | 517.2 GB RAM 775.9 · Disk 672.5 | 542.8 GB RAM 814.1 · Disk 698.5 | 582.0 GB RAM 872.9 · Disk 737.5 |
| FP88bit | 260.4 GB RAM 390.6 · Disk 355.7 | 274.3 GB RAM 411.4 · Disk 381.7 | 301.8 GB RAM 452.7 · Disk 420.7 |
| INT88bit | 260.4 GB RAM 390.6 · Disk 355.7 | 274.3 GB RAM 411.4 · Disk 381.7 | 301.8 GB RAM 452.7 · Disk 420.7 |
| AWQ4bit | 136.8 GB RAM 205.2 · Disk 203.3 | 145.0 GB RAM 217.6 · Disk 229.3 | 166.9 GB RAM 250.4 · Disk 268.3 |
| GPTQ4bit | 138.4 GB RAM 207.6 · Disk 205.3 | 146.7 GB RAM 220.1 · Disk 231.3 | 168.7 GB RAM 253.0 · Disk 270.3 |
| GGUF4bit | 141.6 GB RAM 212.4 · Disk 209.2 | 150.1 GB RAM 225.1 · Disk 235.2 | 172.2 GB RAM 258.3 · Disk 274.2 |
GPU Recommendations
Best ValueBEST
1× B200
180 GB VRAM
$3,013/mo
≈ 0 tok/s (N/A — benchmark unavailable)
Effective production throughput ≈ 0 tok/s
score 0.6124
Cheapest
8× RTX 3090
192 GB VRAM
$636/mo
≈ 28.4 tok/s
Effective production throughput ≈ 19.9 tok/s
score 0.5667
Performance
8× H200 SXM
1128 GB VRAM
$20,440/mo
≈ 145.8 tok/s
Effective production throughput ≈ 102.1 tok/s
score 0.2197
Min Complexity
2× A100 80GB
160 GB VRAM
$1,752/mo
≈ 31 tok/s
Effective production throughput ≈ 21.7 tok/s
score 0.5926
💬 Community — real deployment experience
0 Articles · 0 Benchmarks
GPUs that fit (single card)
B200
180 GB VRAM
$3,013
/mo · vast
B200 SXM
180 GB VRAM
—