Check my GPU →

FLUX.1 Kontext [dev] VRAM requirements

The image-editing member of the FLUX.1 family: give it a picture and an instruction, it changes only what you asked for.

Released 2025-06Licence: FLUX.1 [dev] Non-Commercial LicenseSteps: 20–28Text encoder: T5-XXL (4.9 GB as FP8)
TypeImage
Parameters12B
Fits entirely from7 GB
8-bit or better from15 GB

01Measured on my RTX 5060 Ti 16 GB

FP8 · 1024×1024FASTEST
53.4s / image
flux1-dev-kontext_fp8_scaled.safetensors
Peak VRAM15.0 GB
Run range53.3–53.6 s
Q8_0 · 1024×1024
82.1s / image
flux1-kontext-dev-Q8_0.gguf
Peak VRAM15.5 GB
Run range81.9–82.2 s
FLUX.1 Kontext GGUF Q8_0 output
FLUX.1 Kontext FP8 output
⟷
FP8 · 53.3 sGGUF Q8_0 · 82.1 s
Same seed, same prompt

FP8 or GGUF Q8_0: can you see the difference?

Drag the handle. Both images came out of the same workflow on my RTX 5060 Ti 16 GB; only the model file changed.

FP8 workflow GGUF Q8_0 workflow

02The files, and how much VRAM each needs

FileSizeNeededMin. VRAMQualitySource
16-bit23.8 GB26.4 GB27 GBthe original weightsblack-forest-labs/FLUX.1-Kontext-dev →
Q8_012.7 GB15.3 GB16 GBpractically identical to the originalQuantStack/FLUX.1-Kontext-dev-GGUF →
FP811.9 GB14.5 GB15 GBpractically identical to the originalComfy-Org/flux1-kontext-dev_ComfyUI →
Q6_K9.8 GB12.4 GB13 GBvery close to the originalQuantStack/FLUX.1-Kontext-dev-GGUF →
Q5_K_M8.4 GB11.0 GB12 GBclose; small differences in fine detailQuantStack/FLUX.1-Kontext-dev-GGUF →
Q4_K_M6.9 GB9.5 GB10 GBgood; some loss in fine detail and textQuantStack/FLUX.1-Kontext-dev-GGUF →
Q3_K_M5.4 GB8.0 GB8 GBnoticeable loss of detailQuantStack/FLUX.1-Kontext-dev-GGUF →
Q2_K4.0 GB6.6 GB7 GBheavy loss; a last resortQuantStack/FLUX.1-Kontext-dev-GGUF →

“Needed” = file + 1.8 GB working memory + 0.8 GB system reserve.

Free ComfyUI workflow, tested on my RTX 5060 Ti: FP8 workflow (.json) · GGUF Q8_0 workflow (.json). Drag the file into ComfyUI. All workflows →

03By amount of VRAM

6GBOffload onlyQ3_K_M8GBTightQ3_K_M10GBRunsQ4_K_M12GBRunsQ5_K_M16GBRuns wellQ8_020GBRuns wellQ8_024GBRuns wellQ8_032GBRuns well16-bit

04Best GPU for FLUX.1 Kontext

The cheapest cards (by launch price) that run it well, and every card sorted by memory: best GPU for FLUX.1 Kontext → Planning bigger images or longer clips? Open the calculator →

05By graphics card

GPUVRAMVerdictBest fileNeeded
Desktop graphics cards
RTX 2060 6 GB6 GBOffload onlyQ3_K_M8.0 GB
RTX 3050 6 GB6 GBOffload onlyQ3_K_M8.0 GB
RTX 2070 Super 8 GB8 GBTightQ3_K_M8.0 GB
RTX 2080 Super 8 GB8 GBTightQ3_K_M8.0 GB
RTX 3050 8 GB8 GBTightQ3_K_M8.0 GB
RTX 3060 8 GB8 GBTightQ3_K_M8.0 GB
RTX 3060 Ti 8 GB8 GBTightQ3_K_M8.0 GB
RTX 3070 8 GB8 GBTightQ3_K_M8.0 GB
RTX 3070 Ti 8 GB8 GBTightQ3_K_M8.0 GB
RTX 4060 8 GB8 GBTightQ3_K_M8.0 GB
RTX 4060 Ti 8 GB8 GBTightQ3_K_M8.0 GB
RTX 5050 8 GB8 GBTightQ3_K_M8.0 GB
RTX 5060 8 GB8 GBTightQ3_K_M8.0 GB
RTX 5060 Ti 8 GB8 GBTightQ3_K_M8.0 GB
RX 7600 8 GB8 GBTightQ3_K_M8.0 GB
RX 9050 8 GB8 GBTightQ3_K_M8.0 GB
RX 9060 XT 8 GB8 GBTightQ3_K_M8.0 GB
Arc B570 10 GB10 GBRunsQ4_K_M9.5 GB
RTX 3080 10 GB10 GBRunsQ4_K_M9.5 GB
RTX 2080 Ti 11 GB11 GBRunsQ4_K_M9.5 GB
Arc B580 12 GB12 GBRunsQ5_K_M11.0 GB
RTX 2060 12 GB12 GBRunsQ5_K_M11.0 GB
RTX 3060 12 GB12 GBRunsQ5_K_M11.0 GB
RTX 3080 12 GB12 GBRunsQ5_K_M11.0 GB
RTX 3080 Ti 12 GB12 GBRunsQ5_K_M11.0 GB
RTX 4070 12 GB12 GBRunsQ5_K_M11.0 GB
RTX 4070 Super 12 GB12 GBRunsQ5_K_M11.0 GB
RTX 4070 Ti 12 GB12 GBRunsQ5_K_M11.0 GB
RTX 5070 12 GB12 GBRunsQ5_K_M11.0 GB
RX 7700 XT 12 GB12 GBRunsQ5_K_M11.0 GB
RX 9070 GRE 12 GB12 GBRunsQ5_K_M11.0 GB
Arc A770 16 GB16 GBRuns wellQ8_015.3 GB
RTX 4060 Ti 16 GB16 GBRuns wellFP814.5 GB
RTX 4070 Ti Super 16 GB16 GBRuns wellFP814.5 GB
RTX 4080 16 GB16 GBRuns wellFP814.5 GB
RTX 4080 Super 16 GB16 GBRuns wellFP814.5 GB
RTX 5060 Ti 16 GB16 GBRuns wellFP814.5 GB
RTX 5070 Ti 16 GB16 GBRuns wellFP814.5 GB
RTX 5080 16 GB16 GBRuns wellFP814.5 GB
RX 7600 XT 16 GB16 GBRuns wellQ8_015.3 GB
RX 7800 XT 16 GB16 GBRuns wellQ8_015.3 GB
RX 7900 GRE 16 GB16 GBRuns wellQ8_015.3 GB
RX 9060 XT 16 GB16 GBRuns wellFP814.5 GB
RX 9070 16 GB16 GBRuns wellFP814.5 GB
RX 9070 XT 16 GB16 GBRuns wellFP814.5 GB
RX 7900 XT 20 GB20 GBRuns wellQ8_015.3 GB
Arc Pro B60 24 GB24 GBRuns wellQ8_015.3 GB
RTX 3090 24 GB24 GBRuns wellQ8_015.3 GB
RTX 3090 Ti 24 GB24 GBRuns wellQ8_015.3 GB
RTX 4090 24 GB24 GBRuns wellFP814.5 GB
RX 7900 XTX 24 GB24 GBRuns wellQ8_015.3 GB
Arc Pro B70 32 GB32 GBRuns well16-bit26.4 GB
RTX 5090 32 GB32 GBRuns well16-bit26.4 GB
Laptop GPUs
RTX 3050 Laptop 4 GB4 GBOffload onlyQ4_K_M9.5 GB
RTX 3050 Ti Laptop 4 GB4 GBOffload onlyQ4_K_M9.5 GB
RTX 2060 Laptop 6 GB6 GBOffload onlyQ3_K_M8.0 GB
RTX 3050 Laptop 6 GB6 GBOffload onlyQ3_K_M8.0 GB
RTX 3060 Laptop 6 GB6 GBOffload onlyQ3_K_M8.0 GB
RTX 4050 Laptop 6 GB6 GBOffload onlyQ3_K_M8.0 GB
RTX 2070 Laptop 8 GB8 GBTightQ3_K_M8.0 GB
RTX 2070 Super Laptop 8 GB8 GBTightQ3_K_M8.0 GB
RTX 2080 Laptop 8 GB8 GBTightQ3_K_M8.0 GB
RTX 2080 Super Laptop 8 GB8 GBTightQ3_K_M8.0 GB
RTX 3070 Laptop 8 GB8 GBTightQ3_K_M8.0 GB
RTX 3070 Ti Laptop 8 GB8 GBTightQ3_K_M8.0 GB
RTX 3080 Laptop 8 GB8 GBTightQ3_K_M8.0 GB
RTX 4060 Laptop 8 GB8 GBTightQ3_K_M8.0 GB
RTX 4070 Laptop 8 GB8 GBTightQ3_K_M8.0 GB
RTX 5050 Laptop 8 GB8 GBTightQ3_K_M8.0 GB
RTX 5060 Laptop 8 GB8 GBTightQ3_K_M8.0 GB
RTX 5070 Laptop 8 GB8 GBTightQ3_K_M8.0 GB
RX 7600M 8 GB8 GBTightQ3_K_M8.0 GB
RX 7600M XT 8 GB8 GBTightQ3_K_M8.0 GB
RX 7600S 8 GB8 GBTightQ3_K_M8.0 GB
RX 7700S 8 GB8 GBTightQ3_K_M8.0 GB
RTX 4080 Laptop 12 GB12 GBRunsQ5_K_M11.0 GB
RTX 5070 Laptop 12 GB12 GBRunsQ5_K_M11.0 GB
RTX 5070 Ti Laptop 12 GB12 GBRunsQ5_K_M11.0 GB
RX 7800M 12 GB12 GBRunsQ5_K_M11.0 GB
RTX 3080 Laptop 16 GB16 GBRuns wellQ8_015.3 GB
RTX 3080 Ti Laptop 16 GB16 GBRuns wellQ8_015.3 GB
RTX 4090 Laptop 16 GB16 GBRuns wellFP814.5 GB
RTX 5080 Laptop 16 GB16 GBRuns wellFP814.5 GB
RX 7900M 16 GB16 GBRuns wellQ8_015.3 GB
RTX 5090 Laptop 24 GB24 GBRuns wellFP814.5 GB
Unified memory
Radeon 8060S (Strix Halo) 96 GB96 GBRuns well16-bit26.4 GB
Radeon 8065S (Gorgon Halo) 160 GB160 GBRuns well16-bit26.4 GB
Apple Silicon Macs (by memory)
Mac 16 GB12.7 GBRunsQ6_K12.4 GB
Mac 18 GB14.4 GBRunsQ6_K12.4 GB
Mac 24 GB19.6 GBRuns wellQ8_015.3 GB
Mac 32 GB26.8 GBRuns well16-bit26.4 GB
Mac 36 GB30.2 GBRuns well16-bit26.4 GB
Mac 48 GB40.2 GBRuns well16-bit26.4 GB
Mac 64 GB55.7 GBRuns well16-bit26.4 GB
Mac 96 GB85 GBRuns well16-bit26.4 GB
Mac 128 GB115.4 GBRuns well16-bit26.4 GB
Mac 192 GB175.4 GBRuns well16-bit26.4 GB
Mac 256 GB236.9 GBRuns well16-bit26.4 GB
Mac 512 GB498.1 GBRuns well16-bit26.4 GB

06Text encoder, VAE and other files

FileFolderSizeWhen
CLIP-L
clip_l.safetensors
models/text_encoders0.2 GBrequiredDownload →
T5-XXL FP8
t5xxl_fp8_e4m3fn.safetensors
models/text_encoders4.9 GBtext encoder · smaller, recommendedDownload →
T5-XXL FP16
t5xxl_fp16.safetensors
models/text_encoders9.8 GBtext encoder · alternativeDownload →
FLUX.1 VAE (ae)
ae.safetensors
models/vae0.3 GBrequiredDownload →

The files the official ComfyUI workflows load next to the model. Sizes read from Hugging Face (2026-10-01).

07Where the files go in ComfyUI

FileFolderLoader node
Diffusion model (.safetensors: 16-bit, FP8, INT8)ComfyUI/models/diffusion_modelsLoad Diffusion Model
GGUF file (.gguf)ComfyUI/models/unetUnet Loader (GGUF) - from the ComfyUI-GGUF node pack
Text encoderComfyUI/models/text_encodersLoad CLIP / DualCLIPLoader (or the GGUF versions)
VAEComfyUI/models/vaeLoad VAE

08Measured and reported results

LabelGPUSetupResultPeak VRAMDateSource
measuredRTX 5060 Ti 16 GBflux1-dev-kontext_fp8_scaled.safetensors · 1024x1024 · 20 steps
Image editing: one 1024×1024 input image, 20 steps, euler/simple, guidance 2.5, ReferenceLatent + FluxKontextImageScale. 2 timed runs after 1 warm-up (60.0 s incl. loading). Editin
53.42 s / image15.0 GB2026-10-11
measuredRTX 5060 Ti 16 GBflux1-kontext-dev-Q8_0.gguf · 1024x1024 · 20 steps
Image editing, same settings as the FP8 run: one 1024×1024 input image, 20 steps, euler/simple, guidance 2.5. 2 timed runs after 1 warm-up (94.1 s incl. loading). The GGUF file (Co
82.06 s / image15.5 GB2026-10-11

09Training a LoRA for FLUX.1 Kontext

No trainer documentation with a VRAM figure for this model was found yet. What the trainers say for other models →

10Every file tracked for FLUX.1 Kontext

FileTypeSizeRepo
flux1-dev-kontext_fp8_scaled.safetensorsFP811.90 GBComfy-Org/flux1-kontext-dev_ComfyUI →
flux1-kontext-dev-Q8_0.ggufQ8_012.71 GBQuantStack/FLUX.1-Kontext-dev-GGUF →
flux1-kontext-dev-Q6_K.ggufQ6_K9.85 GBQuantStack/FLUX.1-Kontext-dev-GGUF →
flux1-kontext-dev-Q5_1.ggufQ5_19.02 GBQuantStack/FLUX.1-Kontext-dev-GGUF →
flux1-kontext-dev-Q5_K_M.ggufQ5_K_M8.42 GBQuantStack/FLUX.1-Kontext-dev-GGUF →
flux1-kontext-dev-Q5_0.ggufQ5_08.28 GBQuantStack/FLUX.1-Kontext-dev-GGUF →
flux1-kontext-dev-Q5_K_S.ggufQ5_K_S8.28 GBQuantStack/FLUX.1-Kontext-dev-GGUF →
flux1-kontext-dev-Q4_1.ggufQ4_17.54 GBQuantStack/FLUX.1-Kontext-dev-GGUF →
flux1-kontext-dev-Q4_K_M.ggufQ4_K_M6.93 GBQuantStack/FLUX.1-Kontext-dev-GGUF →
flux1-kontext-dev-Q4_0.ggufQ4_06.80 GBQuantStack/FLUX.1-Kontext-dev-GGUF →
flux1-kontext-dev-Q4_K_S.ggufQ4_K_S6.80 GBQuantStack/FLUX.1-Kontext-dev-GGUF →
flux1-kontext-dev-Q3_K_M.ggufQ3_K_M5.37 GBQuantStack/FLUX.1-Kontext-dev-GGUF →
flux1-kontext-dev-Q3_K_S.ggufQ3_K_S5.23 GBQuantStack/FLUX.1-Kontext-dev-GGUF →
flux1-kontext-dev-Q2_K.ggufQ2_K4.02 GBQuantStack/FLUX.1-Kontext-dev-GGUF →
flux1-kontext-dev.safetensorsBF1623.80 GBblack-forest-labs/FLUX.1-Kontext-dev →

Image editing model. QuantStack Q4_0 and Q4_K_S (and Q5_0/Q5_K_S) have identical byte sizes as listed by HF API.