FLUX.2 [dev] VRAM requirements
The 32B second-generation FLUX. Very capable and very heavy: even 4-bit files are around 20 GB, and the text encoder is a 24B language model.
TypeImage
Parameters32B
Fits entirely from17 GB
8-bit or better from39 GB
01The files, and how much VRAM each needs
| File | Size | Needed | Min. VRAM | Quality | Source |
|---|---|---|---|---|---|
| 16-bit | 64.4 GB | 67.7 GB | 68 GB | the original weights | black-forest-labs/FLUX.2-dev → |
| Q8_0 | 35.0 GB | 38.3 GB | 39 GB | practically identical to the original | city96/FLUX.2-dev-gguf → |
| FP8 | 35.5 GB | 38.8 GB | 39 GB | practically identical to the original | Comfy-Org/flux2-dev → |
| Q6_K | 27.4 GB | 30.7 GB | 31 GB | very close to the original | city96/FLUX.2-dev-gguf → |
| Q5_K_M | 24.1 GB | 27.4 GB | 28 GB | close; small differences in fine detail | city96/FLUX.2-dev-gguf → |
| Q4_K_M | 20.1 GB | 23.4 GB | 24 GB | good; some loss in fine detail and text | city96/FLUX.2-dev-gguf → |
| Q3_K_M | 16.0 GB | 19.3 GB | 20 GB | noticeable loss of detail | city96/FLUX.2-dev-gguf → |
| Q2_K | 12.9 GB | 16.2 GB | 17 GB | heavy loss; a last resort | city96/FLUX.2-dev-gguf → |
“Needed” = file + 2.5 GB working memory + 0.8 GB system reserve.
02By amount of VRAM
03Best GPU for FLUX.2 dev
The cheapest cards (by launch price) that run it well, and every card sorted by memory: best GPU for FLUX.2 dev → Planning bigger images or longer clips? Open the calculator →
04By graphics card
| GPU | VRAM | Verdict | Best file | Needed |
|---|---|---|---|---|
| Desktop graphics cards | ||||
| RTX 2060 6 GB | 6 GB | Not practical | Q2_K | 16.2 GB |
| RTX 3050 6 GB | 6 GB | Not practical | Q2_K | 16.2 GB |
| RTX 2070 Super 8 GB | 8 GB | Not practical | Q2_K | 16.2 GB |
| RTX 2080 Super 8 GB | 8 GB | Not practical | Q2_K | 16.2 GB |
| RTX 3050 8 GB | 8 GB | Not practical | Q2_K | 16.2 GB |
| RTX 3060 8 GB | 8 GB | Not practical | Q2_K | 16.2 GB |
| RTX 3060 Ti 8 GB | 8 GB | Not practical | Q2_K | 16.2 GB |
| RTX 3070 8 GB | 8 GB | Not practical | Q2_K | 16.2 GB |
| RTX 3070 Ti 8 GB | 8 GB | Not practical | Q2_K | 16.2 GB |
| RTX 4060 8 GB | 8 GB | Not practical | Q2_K | 16.2 GB |
| RTX 4060 Ti 8 GB | 8 GB | Not practical | Q2_K | 16.2 GB |
| RTX 5050 8 GB | 8 GB | Not practical | Q2_K | 16.2 GB |
| RTX 5060 8 GB | 8 GB | Not practical | Q2_K | 16.2 GB |
| RTX 5060 Ti 8 GB | 8 GB | Not practical | Q2_K | 16.2 GB |
| RX 7600 8 GB | 8 GB | Not practical | Q2_K | 16.2 GB |
| RX 9050 8 GB | 8 GB | Not practical | Q2_K | 16.2 GB |
| RX 9060 XT 8 GB | 8 GB | Not practical | Q2_K | 16.2 GB |
| Arc B570 10 GB | 10 GB | Not practical | Q2_K | 16.2 GB |
| RTX 3080 10 GB | 10 GB | Not practical | Q2_K | 16.2 GB |
| RTX 2080 Ti 11 GB | 11 GB | Offload only | Q4_K_M | 23.4 GB |
| Arc B580 12 GB | 12 GB | Offload only | Q4_K_M | 23.4 GB |
| RTX 2060 12 GB | 12 GB | Offload only | Q4_K_M | 23.4 GB |
| RTX 3060 12 GB | 12 GB | Offload only | Q4_K_M | 23.4 GB |
| RTX 3080 12 GB | 12 GB | Offload only | Q4_K_M | 23.4 GB |
| RTX 3080 Ti 12 GB | 12 GB | Offload only | Q4_K_M | 23.4 GB |
| RTX 4070 12 GB | 12 GB | Offload only | Q4_K_M | 23.4 GB |
| RTX 4070 Super 12 GB | 12 GB | Offload only | Q4_K_M | 23.4 GB |
| RTX 4070 Ti 12 GB | 12 GB | Offload only | Q4_K_M | 23.4 GB |
| RTX 5070 12 GB | 12 GB | Offload only | Q4_K_M | 23.4 GB |
| RX 7700 XT 12 GB | 12 GB | Offload only | Q4_K_M | 23.4 GB |
| RX 9070 GRE 12 GB | 12 GB | Offload only | Q4_K_M | 23.4 GB |
| Arc A770 16 GB | 16 GB | Offload only | Q2_K | 16.2 GB |
| RTX 4060 Ti 16 GB | 16 GB | Offload only | Q2_K | 16.2 GB |
| RTX 4070 Ti Super 16 GB | 16 GB | Offload only | Q2_K | 16.2 GB |
| RTX 4080 16 GB | 16 GB | Offload only | Q2_K | 16.2 GB |
| RTX 4080 Super 16 GB | 16 GB | Offload only | Q2_K | 16.2 GB |
| RTX 5060 Ti 16 GB | 16 GB | Offload only | Q2_K | 16.2 GB |
| RTX 5070 Ti 16 GB | 16 GB | Offload only | Q2_K | 16.2 GB |
| RTX 5080 16 GB | 16 GB | Offload only | Q2_K | 16.2 GB |
| RX 7600 XT 16 GB | 16 GB | Offload only | Q2_K | 16.2 GB |
| RX 7800 XT 16 GB | 16 GB | Offload only | Q2_K | 16.2 GB |
| RX 7900 GRE 16 GB | 16 GB | Offload only | Q2_K | 16.2 GB |
| RX 9060 XT 16 GB | 16 GB | Offload only | Q2_K | 16.2 GB |
| RX 9070 16 GB | 16 GB | Offload only | Q2_K | 16.2 GB |
| RX 9070 XT 16 GB | 16 GB | Offload only | Q2_K | 16.2 GB |
| RX 7900 XT 20 GB | 20 GB | Tight | Q3_K_M | 19.3 GB |
| Arc Pro B60 24 GB | 24 GB | Runs | Q4_K_M | 23.4 GB |
| RTX 3090 24 GB | 24 GB | Runs | Q4_K_M | 23.4 GB |
| RTX 3090 Ti 24 GB | 24 GB | Runs | Q4_K_M | 23.4 GB |
| RTX 4090 24 GB | 24 GB | Runs | Q4_K_M | 23.4 GB |
| RX 7900 XTX 24 GB | 24 GB | Runs | Q4_K_M | 23.4 GB |
| Arc Pro B70 32 GB | 32 GB | Runs | Q6_K | 30.7 GB |
| RTX 5090 32 GB | 32 GB | Runs | Q6_K | 30.7 GB |
| Laptop GPUs | ||||
| RTX 3050 Laptop 4 GB | 4 GB | Not practical | Q2_K | 16.2 GB |
| RTX 3050 Ti Laptop 4 GB | 4 GB | Not practical | Q2_K | 16.2 GB |
| RTX 2060 Laptop 6 GB | 6 GB | Not practical | Q2_K | 16.2 GB |
| RTX 3050 Laptop 6 GB | 6 GB | Not practical | Q2_K | 16.2 GB |
| RTX 3060 Laptop 6 GB | 6 GB | Not practical | Q2_K | 16.2 GB |
| RTX 4050 Laptop 6 GB | 6 GB | Not practical | Q2_K | 16.2 GB |
| RTX 2070 Laptop 8 GB | 8 GB | Not practical | Q2_K | 16.2 GB |
| RTX 2070 Super Laptop 8 GB | 8 GB | Not practical | Q2_K | 16.2 GB |
| RTX 2080 Laptop 8 GB | 8 GB | Not practical | Q2_K | 16.2 GB |
| RTX 2080 Super Laptop 8 GB | 8 GB | Not practical | Q2_K | 16.2 GB |
| RTX 3070 Laptop 8 GB | 8 GB | Not practical | Q2_K | 16.2 GB |
| RTX 3070 Ti Laptop 8 GB | 8 GB | Not practical | Q2_K | 16.2 GB |
| RTX 3080 Laptop 8 GB | 8 GB | Not practical | Q2_K | 16.2 GB |
| RTX 4060 Laptop 8 GB | 8 GB | Not practical | Q2_K | 16.2 GB |
| RTX 4070 Laptop 8 GB | 8 GB | Not practical | Q2_K | 16.2 GB |
| RTX 5050 Laptop 8 GB | 8 GB | Not practical | Q2_K | 16.2 GB |
| RTX 5060 Laptop 8 GB | 8 GB | Not practical | Q2_K | 16.2 GB |
| RTX 5070 Laptop 8 GB | 8 GB | Not practical | Q2_K | 16.2 GB |
| RX 7600M 8 GB | 8 GB | Not practical | Q2_K | 16.2 GB |
| RX 7600M XT 8 GB | 8 GB | Not practical | Q2_K | 16.2 GB |
| RX 7600S 8 GB | 8 GB | Not practical | Q2_K | 16.2 GB |
| RX 7700S 8 GB | 8 GB | Not practical | Q2_K | 16.2 GB |
| RTX 4080 Laptop 12 GB | 12 GB | Offload only | Q4_K_M | 23.4 GB |
| RTX 5070 Laptop 12 GB | 12 GB | Offload only | Q4_K_M | 23.4 GB |
| RTX 5070 Ti Laptop 12 GB | 12 GB | Offload only | Q4_K_M | 23.4 GB |
| RX 7800M 12 GB | 12 GB | Offload only | Q4_K_M | 23.4 GB |
| RTX 3080 Laptop 16 GB | 16 GB | Offload only | Q2_K | 16.2 GB |
| RTX 3080 Ti Laptop 16 GB | 16 GB | Offload only | Q2_K | 16.2 GB |
| RTX 4090 Laptop 16 GB | 16 GB | Offload only | Q2_K | 16.2 GB |
| RTX 5080 Laptop 16 GB | 16 GB | Offload only | Q2_K | 16.2 GB |
| RX 7900M 16 GB | 16 GB | Offload only | Q2_K | 16.2 GB |
| RTX 5090 Laptop 24 GB | 24 GB | Runs | Q4_K_M | 23.4 GB |
| Unified memory | ||||
| Radeon 8060S (Strix Halo) 96 GB | 96 GB | Runs well | 16-bit | 67.7 GB |
| Radeon 8065S (Gorgon Halo) 160 GB | 160 GB | Runs well | 16-bit | 67.7 GB |
| Apple Silicon Macs (by memory) | ||||
| Mac 16 GB | 12.7 GB | Offload only | Q4_K_M | 23.4 GB |
| Mac 18 GB | 14.4 GB | Offload only | Q2_K | 16.2 GB |
| Mac 24 GB | 19.6 GB | Tight | Q3_K_M | 19.3 GB |
| Mac 32 GB | 26.8 GB | Runs | Q4_K_M | 23.4 GB |
| Mac 36 GB | 30.2 GB | Runs | Q5_K_M | 27.4 GB |
| Mac 48 GB | 40.2 GB | Runs well | Q8_0 | 38.3 GB |
| Mac 64 GB | 55.7 GB | Runs well | Q8_0 | 38.3 GB |
| Mac 96 GB | 85 GB | Runs well | 16-bit | 67.7 GB |
| Mac 128 GB | 115.4 GB | Runs well | 16-bit | 67.7 GB |
| Mac 192 GB | 175.4 GB | Runs well | 16-bit | 67.7 GB |
| Mac 256 GB | 236.9 GB | Runs well | 16-bit | 67.7 GB |
| Mac 512 GB | 498.1 GB | Runs well | 16-bit | 67.7 GB |
05Text encoder, VAE and other files
| File | Folder | Size | When | |
|---|---|---|---|---|
| Mistral 3 Small FP8 mistral_3_small_flux2_fp8.safetensors | models/text_encoders | 18.0 GB | text encoder · smaller, recommended | Download → |
| Mistral 3 Small BF16 mistral_3_small_flux2_bf16.safetensors | models/text_encoders | 35.6 GB | text encoder · alternative | Download → |
| FLUX.2 VAE flux2-vae.safetensors | models/vae | 0.3 GB | required | Download → |
The files the official ComfyUI workflows load next to the model. Sizes read from Hugging Face (2026-10-01).
At 18.0 GB even the smallest Mistral Small 3 (24B) file is too big for 16 GB, so it runs from system RAM. That slows down reading the prompt, not the picture. Sizes: 16-bit 35.6 GB, FP8 18.0 GB. Why the encoder does not count against VRAM →
06Where the files go in ComfyUI
| File | Folder | Loader node |
|---|---|---|
| Diffusion model (.safetensors: 16-bit, FP8, INT8) | ComfyUI/models/diffusion_models | Load Diffusion Model |
| GGUF file (.gguf) | ComfyUI/models/unet | Unet Loader (GGUF) - from the ComfyUI-GGUF node pack |
| Text encoder | ComfyUI/models/text_encoders | Load CLIP / DualCLIPLoader (or the GGUF versions) |
| VAE | ComfyUI/models/vae | Load VAE |
07Measured and reported results
| Label | GPU | Setup | Result | Peak VRAM | Date | Source |
|---|---|---|---|---|---|---|
| reported | RTX 5060 Ti 16 GB | flux2_dev_fp8mixed + mistral_3_small_flux2_fp8 · 1024x1024 · 20 steps “FP8 t2i 142秒程度（20Steps）” ComfyUI t2i at 1MP; approximate (程度) | 142 s / image | — | 2025-11-28 | note.com → |
| reported | RTX 5060 Ti 16 GB | GGUF Q4_0 (flux2-dev-q4_0.gguf) · 1024x1024 · 20 steps “GGUF Q4_0 t2i 180秒程度（20Steps）” ComfyUI t2i at 1MP; approximate (程度); author notes GGUF slower than fp8 when spilling VRAM | 180 s / image | — | 2025-11-28 | note.com → |
Other people's numbers are copied as published, with a link. Send yours.
08Training a LoRA for FLUX.2 dev
No trainer documentation with a VRAM figure for this model was found yet. What the trainers say for other models →
09Every file tracked for FLUX.2 dev
| File | Type | Size | Repo |
|---|---|---|---|
| flux2_dev_fp8mixed.safetensors | FP8 | 35.46 GB | Comfy-Org/flux2-dev → |
| flux2-dev.safetensors | BF16 | 64.45 GB | black-forest-labs/FLUX.2-dev → |
| flux2-dev-BF16.gguf | BF16 | 64.45 GB | city96/FLUX.2-dev-gguf → |
| flux2-dev-Q8_0.gguf | Q8_0 | 35.00 GB | city96/FLUX.2-dev-gguf → |
| flux2-dev-Q6_K.gguf | Q6_K | 27.40 GB | city96/FLUX.2-dev-gguf → |
| flux2-dev-Q5_1.gguf | Q5_1 | 25.19 GB | city96/FLUX.2-dev-gguf → |
| flux2-dev-Q5_K_M.gguf | Q5_K_M | 24.06 GB | city96/FLUX.2-dev-gguf → |
| flux2-dev-Q5_0.gguf | Q5_0 | 23.22 GB | city96/FLUX.2-dev-gguf → |
| flux2-dev-Q5_K_S.gguf | Q5_K_S | 23.22 GB | city96/FLUX.2-dev-gguf → |
| flux2-dev-Q4_1.gguf | Q4_1 | 21.26 GB | city96/FLUX.2-dev-gguf → |
| flux2-dev-Q4_K_M.gguf | Q4_K_M | 20.08 GB | city96/FLUX.2-dev-gguf → |
| flux2-dev-Q4_0.gguf | Q4_0 | 19.30 GB | city96/FLUX.2-dev-gguf → |
| flux2-dev-Q4_K_S.gguf | Q4_K_S | 19.30 GB | city96/FLUX.2-dev-gguf → |
| flux2-dev-Q3_K_M.gguf | Q3_K_M | 15.96 GB | city96/FLUX.2-dev-gguf → |
| flux2-dev-Q3_K_S.gguf | Q3_K_S | 15.78 GB | city96/FLUX.2-dev-gguf → |
| flux2-dev-Q2_K.gguf | Q2_K | 12.86 GB | city96/FLUX.2-dev-gguf → |
Text encoder is Mistral Small 3 (24B) - see mistral-3-small-flux2 entry; very heavy. VAE ae.safetensors 336211292 bytes (BFL repo).