Qwen3-Coder 480B
Qwen/Qwen3-Coder-480B-A35B-Instruct
Parameters
480B
Context
256K
Downloads
88.7K
Likes
1.4K
Architecture
moe_transformer
GPU Requirements
VRAM / RAM / disk estimates by quantization and usage scenario
| Quantization | Minimum | Recommended | Production |
|---|---|---|---|
| FP1616bit | 1136.0 GB RAM 1703.9 · Disk 1433.8 | 1203.1 GB RAM 1804.7 · Disk 1459.8 | 1351.4 GB RAM 2027.1 · Disk 1498.8 |
| FP88bit | 570.5 GB RAM 855.7 · Disk 736.4 | 611.9 GB RAM 917.9 · Disk 762.4 | 734.5 GB RAM 1101.8 · Disk 801.4 |
| INT88bit | 570.5 GB RAM 855.7 · Disk 736.4 | 611.9 GB RAM 917.9 · Disk 762.4 | 734.5 GB RAM 1101.8 · Disk 801.4 |
| AWQ4bit | 298.3 GB RAM 447.5 · Disk 400.8 | 327.4 GB RAM 491.1 · Disk 426.8 | 437.6 GB RAM 656.4 · Disk 465.8 |
| GPTQ4bit | 301.8 GB RAM 452.8 · Disk 405.1 | 331.1 GB RAM 496.6 · Disk 431.1 | 441.5 GB RAM 662.2 · Disk 470.1 |
| GGUF4bit | 308.9 GB RAM 463.4 · Disk 413.8 | 338.5 GB RAM 507.7 · Disk 439.8 | 449.2 GB RAM 673.8 · Disk 478.8 |
GPU Recommendations
Best ValueBEST
8× H200 SXM
1128 GB VRAM
$20,440/mo
≈ 66.2 tok/s
Effective production throughput ≈ 46.3 tok/s
score 0.2971
Cheapest
8× RTX A6000
384 GB VRAM
$1,678/mo
≈ 10.6 tok/s
Effective production throughput ≈ 7.4 tok/s
score 0.5152
Performance
4× H200 SXM
564 GB VRAM
$10,220/mo
≈ 49.7 tok/s
Effective production throughput ≈ 34.8 tok/s
score 0.4242
Min Complexity
2× B200
360 GB VRAM
$6,026/mo
≈ 0 tok/s (N/A — benchmark unavailable)
Effective production throughput ≈ 0 tok/s
score 0.4795
💬 Community — real deployment experience
0 Articles · 0 Benchmarks