How much VRAM does FLUX.2 [klein] 4B need?

FLUX.2 [klein] 4B is an image generation model by Black Forest Labs (3.9B; text encoder: Qwen3-4B). Fast 4-step model that also does image editing; officially about 13 GB of VRAM.

5.2 GBVRAM needed at FP8, 1024×1024 (with model swapping)

Change precision, resolution →

VRAM by precision and resolution

Precision	Model size	1024×1024	1536×1536
BF16 / FP16 (uncompressed)	7.8 GB	9.0 GB all in VRAM 13.6 GB	9.8 GB all in VRAM 14.4 GB
FP8 (half size, near-lossless)	4.0 GB	5.2 GB all in VRAM 9.8 GB	6.0 GB all in VRAM 10.6 GB
GGUF Q8_0 (similar to FP8)	4.2 GB	5.4 GB all in VRAM 10.0 GB	6.2 GB all in VRAM 10.8 GB
GGUF Q4_K_S (popular low-VRAM choice)	2.3 GB	4.6 GB all in VRAM 8.1 GB	4.6 GB all in VRAM 9.0 GB

Bold = VRAM needed when ComfyUI swaps the text encoder and the model. "All in VRAM" = keeping both loaded. Text encoder assumed at FP8. Official: runs in about 13 GB of VRAM (RTX 3090 / 4070 and up) (measured with the official implementation).

GPUs that run FLUX.2 [klein] 4B (34 GPUs, 1024×1024)

GPU	VRAM	Best precision	VRAM needed	Verdict	Buy
RTX 4060 previous gen	8 GB	FP8	5.2 GB	○ Runs with model swapping	Check price
RTX 4060 Ti 8GB previous gen	8 GB	FP8	5.2 GB	○ Runs with model swapping	Check price
Radeon RX 7600 previous gen	8 GB	FP8	5.2 GB	○ Runs with model swapping	Check price
RTX 5050	8 GB	FP8	5.2 GB	○ Runs with model swapping	Check price
RTX 5060 Ti 8GB	8 GB	FP8	5.2 GB	○ Runs with model swapping	Check price
RTX 5060	8 GB	FP8	5.2 GB	○ Runs with model swapping	Check price
RTX 3070 previous gen	8 GB	FP8	5.2 GB	○ Runs with model swapping	Check price
RTX 3060 Ti previous gen	8 GB	FP8	5.2 GB	○ Runs with model swapping	Check price
RTX 3080 (10GB) previous gen	10 GB	BF16 / FP16	9.0 GB	○ Runs with model swapping	Check price
RTX 2080 Ti previous gen	11 GB	BF16 / FP16	9.0 GB	○ Runs with model swapping	Check price
RTX 3060 12GB previous gen	12 GB	BF16 / FP16	9.0 GB	○ Runs with model swapping	Check price
Radeon RX 6700 XT previous gen	12 GB	BF16 / FP16	9.0 GB	○ Runs with model swapping	Check price
Intel Arc B580	12 GB	BF16 / FP16	9.0 GB	○ Runs with model swapping	Check price
RTX 4070 Ti previous gen	12 GB	BF16 / FP16	9.0 GB	○ Runs with model swapping	Check price
RTX 4070 / SUPER previous gen	12 GB	BF16 / FP16	9.0 GB	○ Runs with model swapping	Check price
RTX 5070	12 GB	BF16 / FP16	9.0 GB	○ Runs with model swapping	Check price
RTX 4060 Ti 16GB previous gen	16 GB	BF16 / FP16	9.0 GB	◎ Fits entirely in VRAM	Check price
Radeon RX 7600 XT previous gen	16 GB	BF16 / FP16	9.0 GB	◎ Fits entirely in VRAM	Check price
Radeon RX 9060 XT 16GB	16 GB	BF16 / FP16	9.0 GB	◎ Fits entirely in VRAM	Check price
RTX 5060 Ti 16GB	16 GB	BF16 / FP16	9.0 GB	◎ Fits entirely in VRAM	Check price
Radeon RX 6800 previous gen	16 GB	BF16 / FP16	9.0 GB	◎ Fits entirely in VRAM	Check price
Radeon RX 7800 XT previous gen	16 GB	BF16 / FP16	9.0 GB	◎ Fits entirely in VRAM	Check price
Radeon RX 9070 XT	16 GB	BF16 / FP16	9.0 GB	◎ Fits entirely in VRAM	Check price
Radeon RX 9070	16 GB	BF16 / FP16	9.0 GB	◎ Fits entirely in VRAM	Check price
RTX 4070 Ti SUPER previous gen	16 GB	BF16 / FP16	9.0 GB	◎ Fits entirely in VRAM	Check price
RTX 4080 previous gen	16 GB	BF16 / FP16	9.0 GB	◎ Fits entirely in VRAM	Check price
RTX 4080 SUPER previous gen	16 GB	BF16 / FP16	9.0 GB	◎ Fits entirely in VRAM	Check price
RTX 5070 Ti	16 GB	BF16 / FP16	9.0 GB	◎ Fits entirely in VRAM	Check price
RTX 5080	16 GB	BF16 / FP16	9.0 GB	◎ Fits entirely in VRAM	Check price
Radeon RX 7900 XT previous gen	20 GB	BF16 / FP16	9.0 GB	◎ Fits entirely in VRAM	Check price
RTX 3090 previous gen	24 GB	BF16 / FP16	9.0 GB	◎ Fits entirely in VRAM	Check price
Radeon RX 7900 XTX previous gen	24 GB	BF16 / FP16	9.0 GB	◎ Fits entirely in VRAM	Check price
RTX 4090 previous gen	24 GB	BF16 / FP16	9.0 GB	◎ Fits entirely in VRAM	Check price
RTX 5090	32 GB	BF16 / FP16	9.0 GB	◎ Fits entirely in VRAM	Check price

ComfyUI also runs on AMD Radeon (ROCm) and Intel Arc, but setup is harder than on NVIDIA and some features (such as FP8 acceleration) are not supported.

License

License: Apache 2.0 (commercial use: Allowed). Always read the full license before using outputs commercially.
Download: Hugging Face (black-forest-labs/FLUX.2-klein-4B). FP8 and GGUF builds (loaded with ComfyUI-GGUF) are widely shared.

Other image & video models

Stable Diffusion 1.5Image · 0.86B SDXL 1.0 (incl. Illustrious, Pony and other fine-tunes)Image · 2.57B Stable Diffusion 3.5 LargeImage · 8.1B FLUX.1 [dev]Image · 11.9B FLUX.2 [klein] 9BImage · 9B FLUX.2 [dev]Image · 32B Qwen-Image (incl. 2512 and Edit)Image · 20.4B Z-Image TurboImage · 6.15B HiDream-I1Image · 17.1B Wan 2.1 T2V 1.3BVideo · 1.42B Wan 2.2 TI2V 5BVideo · 5B Wan 2.2 A14B (T2V / I2V)Video · 14.3B×2 HunyuanVideo 1.5Video · 8.3B LTX-2.3 (22B, with audio)Video · 22B

→ Image & Video AI VRAM Calculator (all 15 models) / Local LLM VRAM Calculator

Estimates for ComfyUI, about ±20%. Model specs from the publisher's config files and released file sizes (checked October 2026).