Qwen/Qwen3.6-35B-A3B-FP8
Qwen/Qwen3.6-35B-A3B-FP8
Parameters
2.9B
Context
256K
Downloads
10.7M
Likes
351
Architecture
qwen3_5_moe
40 layers · 2 KV heads · head 256 · hidden 2048
GPU Requirements
VRAM / RAM / disk estimates by quantization and usage scenario
| Quantization | Minimum | Recommended | Production |
|---|---|---|---|
| FP1616bit | 10.8 GB RAM 32.0 · Disk 47.4 | 21.0 GB RAM 32.0 · Disk 73.4 | 81.9 GB RAM 122.8 · Disk 112.4 |
| FP88bit | 7.3 GB RAM 32.0 · Disk 43.2 | 17.4 GB RAM 32.0 · Disk 69.2 | 78.1 GB RAM 117.2 · Disk 108.2 |
| INT88bit | 7.3 GB RAM 32.0 · Disk 43.2 | 17.4 GB RAM 32.0 · Disk 69.2 | 78.1 GB RAM 117.2 · Disk 108.2 |
| AWQ4bit | 5.7 GB RAM 32.0 · Disk 41.2 | 15.7 GB RAM 32.0 · Disk 67.2 | 76.3 GB RAM 114.5 · Disk 106.2 |
| GPTQ4bit | 5.7 GB RAM 32.0 · Disk 41.2 | 15.7 GB RAM 32.0 · Disk 67.2 | 76.4 GB RAM 114.5 · Disk 106.2 |
| GGUF4bit | 5.8 GB RAM 32.0 · Disk 41.3 | 15.7 GB RAM 32.0 · Disk 67.3 | 76.4 GB RAM 114.6 · Disk 106.3 |
GPU Recommendations
1× RTX 5090
32 GB VRAM
≈ 1280 tok/s
Effective production throughput ≈ 896 tok/s
score 0.6286
1× RTX 4070S Ti
16 GB VRAM
≈ 0 tok/s (N/A — benchmark unavailable)
Effective production throughput ≈ 0 tok/s
score 0.6356
8× H200 SXM
1128 GB VRAM
≈ 10971.4 tok/s
Effective production throughput ≈ 7680 tok/s
score 0.2197
1× RTX 3090
24 GB VRAM
≈ 668.6 tok/s
Effective production throughput ≈ 468 tok/s
score 0.6691
💬 Community — real deployment experience
0 Articles · 0 Benchmarks
GPUs that fit (single card)
RTX 4070S Ti
16 GB VRAM
$49.5
/mo · vast
Tesla V100
16 GB VRAM
$58.0
/mo · vast
RTX 4080S
16 GB VRAM
$64.6
/mo · vast
RTX 3090
24 GB VRAM
$78.3
/mo · vast
RTX 5080
16 GB VRAM
$85.6
/mo · vast