How much VRAM does FLUX.3 Klein need?
FLUX.3 Klein needs about 6.4 GB at 8-bit, or 4.1 GB as a 4-bit GGUF, at its default output size.
- Type
- Image model
- Parameters
- 5.5B
- Text encoder
- Qwen3-VL-4B-Instruct
- Released
- Oct 2026
- License
- FLUX Kommunity License v1.0
Memory at each precision
| Precision | Weights | Peak at 1024 × 1024 |
|---|---|---|
| 16-bit BF16 | 10.3 GB | 11.3 GB |
| 8-bit FP8 · Q8_0 | 5.4 GB | 6.4 GB |
| 6-bit GGUF Q6_K | 4.3 GB | 5.3 GB |
| 5-bit GGUF Q5_K_M | 3.7 GB | 4.7 GB |
| 4-bit GGUF Q4_K_M · NVFP4 | 3.1 GB | 4.1 GB |
The Qwen3-VL-4B-Instruct text encoder (4.1 GB at FP8) runs first and then waits in system RAM.
What runs it
| Machine | Best fit | Per image |
|---|---|---|
| RTX 3060 12 GB | 16-bit | 52 s |
| RTX 4060 Ti 16 GB | 16-bit | 31 s |
| RTX 3090 | 16-bit | 19 s |
| RTX 4090 | 16-bit | 8.9 s |
| RTX 5090 | 16-bit | 7.2 s |
| RTX PRO 6000 Blackwell | 16-bit | 3.6 s |
| M4 Max 128 GB | 16-bit | 86 s |
| M5 Max 128 GB | 16-bit | 26 s |
| Ryzen AI Max+ 395 128 GB | 16-bit | 39 s |
| DGX Spark 128 GB | 16-bit | 22 s |
| M3 Ultra 256 GB | 16-bit | 49 s |
| M5 Ultra 512 GB | 16-bit | 16 s |
| H100 SXM | 16-bit | 2.3 s |
| DGX Station GB300 | 16-bit | 1.5 s |
1024 × 1024, 16 steps. Times are rough, plus or minus 30%.
Where to get it
Questions
How much VRAM does FLUX.3 Klein need?
FLUX.3 Klein needs about 6.4 GB at 8-bit, or 4.1 GB as a 4-bit GGUF, at its default output size.
Can FLUX.3 Klein run on an RTX 4090?
Yes. At 16-bit it takes about 8.9 s per image.