Alibaba
Qwen3.5
MoE
Pick a size: the video memory shown is the Q4_K_M file (the protocol's quantisation) plus about 1.5 GB for a 4,000-token context.
0.8B
8 GB card- Parameters
- 0,9B
- Q4_K_M file
- 0,5 Go
- Recommended VRAM
- ≥ 2,0 Go
- Max context
- 262 144
Not measured yet
2B
8 GB card- Parameters
- 2,3B
- Q4_K_M file
- 1,3 Go
- Recommended VRAM
- ≥ 2,8 Go
- Max context
- 262 144
Not measured yet
4B
8 GB card- Parameters
- 4,7B
- Q4_K_M file
- 2,8 Go
- Recommended VRAM
- ≥ 4,3 Go
- Max context
- 262 144
Not measured yet
9B
8 GB card- Parameters
- 9,7B
- Q4_K_M file
- 5,7 Go
- Recommended VRAM
- ≥ 7,2 Go
- Max context
- 262 144
Not measured yet
27B
24 GB card- Parameters
- 27,8B
- Q4_K_M file
- 16,7 Go
- Recommended VRAM
- ≥ 18,2 Go
- Max context
- 262 144
Not measured yet
35B-A3B
24 GB card- Parameters
- 36B · 3B active
- Q4_K_M file
- 20,8 Go
- Recommended VRAM
- ≥ 22,3 Go
- Max context
- 262 144
Not measured yet
122B-A10B
> 32 GB- Parameters
- 125,1B · 10B active
- Q4_K_M file
- 72,3 Go
- Recommended VRAM
- ≥ 73,8 Go
- Max context
- 262 144
Not measured yet
397B-A17B
> 32 GB- Parameters
- 403,4B · 17B active
- Q4_K_M file
- 231,2 Go
- Recommended VRAM
- ≥ 232,7 Go
- Max context
- 262 144
Not measured yet
MoE (“30B-A3B”): all the memory of a 30B, but the speed of a model with 3B active parameters. “> 32 GB”: several cards, or part of the model in system memory (much slower).