Black Forest LabsDiffusion TransformerNot for commercial use
FLUX.2 [klein] 9B
Better quality than klein 4B at roughly twice the weights, but it moves to the non-commercial licence. Gated on Hugging Face.
Parameters9.08BImage model
Weights at BF1634.7 GBAll components
Native resolutionup to 4MPTrained for
Minimum VRAM8.4 GBNF4 + text encoder on CPU
What you actually download
A diffusion pipeline is three separate models. These sizes are read from the real .safetensors files in the Hugging Face repository, not estimated.
Transformer / UNetThe image model itself18.16 GB
Text encoderTurns your prompt into embeddings16.38 GB
VAEDecodes latents into pixels0.17 GB
Peak VRAM at 1024×1024, one image
Every combination of precision and offload strategy. Lower precision costs image quality; offloading costs speed.
| Precision | Everything in VRAM | Text encoder on CPU | Sequential CPU offload |
|---|---|---|---|
| BF16 / FP16Reference quality. What the repos ship by default. | 37.5 GB | 21.1 GB | 21.0 GB |
| FP8Roughly half the weight memory. Small quality cost, widely used for Flux and SD 3.5. | 20.2 GB | 12.1 GB | 11.9 GB |
| NF4 / INT4Aggressive 4-bit. Visible quality loss, but it puts big models on small cards. | 16.6 GB | 8.4 GB | 11.0 GB |
Which machines can run it
64 of the 75 devices we profile can run FLUX.2 [klein] 9B at 1024×1024 in some configuration. Showing the best-quality setup each one supports.
| Device | Best setup | Peak | Status |
|---|---|---|---|
| Apple Mac Studio (M5 Ultra, 512GB, 2026) | BF16 / FP16 | 37.5 GB | Comfortable |
| Apple Mac Studio (M3 Ultra, 512GB, 2025) | BF16 / FP16 | 37.5 GB | Comfortable |
| Apple Mac Studio (M5 Ultra, 256GB, 2026) | BF16 / FP16 | 37.5 GB | Comfortable |
| Apple Mac Studio (M2 Ultra 192GB) | BF16 / FP16 | 37.5 GB | Comfortable |
| NVIDIA DGX Spark (GB10, 128GB) | BF16 / FP16 | 37.5 GB | Comfortable |
| Apple MacBook Pro 16-inch (M4 Max 128GB) | BF16 / FP16 | 37.5 GB | Comfortable |
| Apple MacBook Pro 16-inch (M3 Max 128GB) | BF16 / FP16 | 37.5 GB | Comfortable |
| Next-Gen AI Laptop (Ryzen AI Max+ 395 128GB) | BF16 / FP16 | 37.5 GB | Comfortable |
| Apple Mac Studio (M5 Max, 128GB, 2026) | BF16 / FP16 | 37.5 GB | Comfortable |
| Apple MacBook Pro 16-inch (M4 Max) | BF16 / FP16 | 37.5 GB | Comfortable |
| Apple Mac Studio (M1 Ultra 64GB) | BF16 / FP16 | 37.5 GB | Comfortable |
| Homelab AI Server (Dual RTX 3090 48GB VRAM) | BF16 / FP16 | 37.5 GB | Comfortable |
| Apple Mac mini (M5 Pro, 64GB, 2026) | BF16 / FP16 | 37.5 GB | Comfortable |
| Framework Laptop 13 (Ryzen 7 7840U) | BF16 / FP16 | 37.5 GB | Tight Fit |
| Apple MacBook Pro 14-inch (M4 Pro 48GB) | BF16 / FP16Text encoder on CPU | 21.1 GB | With offload |
| Apple Mac mini (M5 Pro, 48GB, 2026) | BF16 / FP16Text encoder on CPU | 21.1 GB | With offload |
| Flagship AI Workstation (RTX 5090 32GB) | BF16 / FP16Text encoder on CPU | 21.1 GB | With offload |
| ASUS Zenbook A16 (Snapdragon X2 Elite Extreme, 48GB) | BF16 / FP16Text encoder on CPU | 21.1 GB | With offload |
| Apple MacBook Pro 14-inch (M3 Pro 36GB) | BF16 / FP16Text encoder on CPU | 21.1 GB | With offload |
| Custom Workstation PC (RTX 4090 24GB) | BF16 / FP16Text encoder on CPU | 21.1 GB | With offload |
| Used AI Powerhouse PC (RTX 3090 24GB) | BF16 / FP16Text encoder on CPU | 21.1 GB | With offload |
| Custom Workstation PC (Radeon RX 7900 XTX 24GB) | BF16 / FP16Text encoder on CPU | 21.1 GB | With offload |
| Apple Mac mini (M6, 32GB, 2026) | BF16 / FP16Text encoder on CPU | 21.1 GB | With offload |
| Apple MacBook Pro 14-inch (M5, 32GB, 2025) | BF16 / FP16Text encoder on CPU | 21.1 GB | With offload |
| HP Omnibook 5 OLED | FP8Text encoder on CPU | 12.1 GB | With offload |
| Lenovo ThinkPad T14s Gen 5 (Snapdragon) | FP8Text encoder on CPU | 12.1 GB | With offload |
| ASUS Zenbook S 16 (Ryzen AI 9 HX 370) | FP8Text encoder on CPU | 12.1 GB | With offload |
| Dell XPS 13 (Intel Core Ultra 7 258V) | FP8Text encoder on CPU | 12.1 GB | With offload |
| Lenovo Yoga Slim 7x Copilot+ | FP8Text encoder on CPU | 12.1 GB | With offload |
| Microsoft Surface Pro 11 Copilot+ (OLED) | FP8Text encoder on CPU | 12.1 GB | With offload |
Showing the 30 devices with the most memory. The full hardware matrix covers all 75.
Licence: FLUX Non-Commercial License
No. FLUX.2 [klein] 9B is released under FLUX Non-Commercial License, which does not permit commercial use. You would need a separate agreement with the rights holder to ship output commercially.
Compare model licences →FLUX.2 [klein] 9B: common questions
- How much VRAM does FLUX.2 [klein] 9B need?
- At 1024x1024 and batch 1, FLUX.2 [klein] 9B peaks at about 37.5GB with everything resident at BF16, and as little as 8.4GB at NF4 + text encoder on CPU. The weights alone are 34.7GB at BF16 — 18.2GB of transformer, 16.4GB of text encoder and 0.17GB of VAE.
- Which GPUs can run FLUX.2 [klein] 9B?
- 64 of the 75 devices we profile can run it at 1024x1024 in some configuration. The Apple Mac Studio (M5 Ultra, 512GB, 2026) handles it best, at about 37.5GB. Lower precision and text-encoder offload widen that list considerably.
- Can FLUX.2 [klein] 9B be used commercially?
- No. FLUX.2 [klein] 9B is released under FLUX Non-Commercial License, which does not permit commercial use. You would need a separate agreement with the rights holder to ship output commercially.
- Does offloading the text encoder help on FLUX.2 [klein] 9B?
- Yes, substantially. FLUX.2 [klein] 9B uses a large language model as its text encoder — 16.4GB at full precision. Encoding happens once per prompt, so moving it to CPU costs a couple of seconds and frees that 16.4GB for the rest of the run, taking peak VRAM at BF16 from 37.5GB down to 21.1GB.
Weight sizes read from the model repository on 2026-09-12. Activation memory is an approximation, not a benchmark — BFWR does not run a hardware test lab.
Other open image models
Stable Diffusion 1.50.86B · 2.1GB at BF16 · conditions applyStable Diffusion XL 1.02.57B · 6.9GB at BF16 · conditions applyStable Diffusion 3.5 Medium2.47B · 16.5GB at BF16 · conditions applyStable Diffusion 3.5 Large8.15B · 27.9GB at BF16 · conditions applyFLUX.1 [schnell]11.89B · 33.7GB at BF16 · commercial OKFLUX.1 [dev]11.9B · 33.7GB at BF16 · non-commercialQwen-Image20.43B · 57.7GB at BF16 · commercial OKQwen-Image 2.17.12B · 32.4GB at BF16 · non-commercialFLUX.2 [klein] 4B3.88B · 16.0GB at BF16 · commercial OKFLUX.2 [dev]32.22B · 112.8GB at BF16 · non-commercial
Guides that go deeper
Which GPUs have the VRAM these models need, and what they cost right now.
- How to Run Stable Diffusion Locally: Hardware, VRAM and SetupCurrent models from SD 1.5 to Qwen-Image, and the VRAM each needs.Read the guide
- Flux LoRA: Which Base Model You Can Legally Train OnA LoRA is a Derivative; only klein 4B and schnell are Apache-2.0.Read the guide
- How to Increase VRAM: What Actually Works on PC, Laptop and MacIt depends where the memory physically lives.Read the guide