How much VRAM does FLUX.2 [klein] 9B need?
FLUX.2 [klein] 9B needs about 9.9 GB at 8-bit, or 6.2 GB as a 4-bit GGUF, at its default output size.
- Type
- Image model
- Parameters
- 9.1B
- Text encoder
- Qwen3-8B
- Released
- Jan 2026
- License
- FLUX Non-Commercial License
Memory at each precision
| Precision | Weights | Peak at 1024 × 1024 |
|---|---|---|
| 16-bit BF16 | 16.9 GB | 18.0 GB |
| 8-bit FP8 · Q8_0 | 8.8 GB | 9.9 GB |
| 6-bit GGUF Q6_K | 7.0 GB | 8.1 GB |
| 5-bit GGUF Q5_K_M | 6.0 GB | 7.1 GB |
| 4-bit GGUF Q4_K_M · NVFP4 | 5.1 GB | 6.2 GB |
The Qwen3-8B text encoder (7.6 GB at FP8) runs first and then waits in system RAM.
What runs it
| Machine | Best fit | Per image |
|---|---|---|
| RTX 3060 12 GB | 8-bit | 17 s |
| RTX 4060 Ti 16 GB | 8-bit | 10 s |
| RTX 3090 | 16-bit | 6.8 s |
| RTX 4090 | 16-bit | 3.5 s |
| RTX 5090 | 16-bit | 3.0 s |
| RTX PRO 6000 Blackwell | 16-bit | 1.8 s |
| M4 Max 128 GB | 16-bit | 28 s |
| M5 Max 128 GB | 16-bit | 8.8 s |
| Ryzen AI Max+ 395 128 GB | 16-bit | 13 s |
| DGX Spark 128 GB | 16-bit | 7.6 s |
| M3 Ultra 256 GB | 16-bit | 16 s |
| M5 Ultra 512 GB | 16-bit | 5.7 s |
| H100 SXM | 16-bit | 1.4 s |
| DGX Station GB300 | 16-bit | 1.2 s |
1024 × 1024, 4 steps. Times are rough, plus or minus 30%.
Where to get it
Questions
How much VRAM does FLUX.2 [klein] 9B need?
FLUX.2 [klein] 9B needs about 9.9 GB at 8-bit, or 6.2 GB as a 4-bit GGUF, at its default output size.
Can FLUX.2 [klein] 9B run on an RTX 4090?
Yes. At 16-bit it takes about 3.5 s per image.