Check my GPU →

HunyuanImage 2.1 VRAM requirements

Tencent's 17B model that generates natively at 2K. Its GGUF files come from calcuis and need calcuis's own GGUF loader node.

Released 2025-09Licence: Tencent Hunyuan Community License (excludes EU/UK/KR)Steps: 20–50Text encoder: Qwen2.5-VL 7B (9.4 GB as FP8)
TypeImage
Parameters17B
Fits entirely from11 GB
8-bit or better from21 GB

01The files, and how much VRAM each needs

FileSizeNeededMin. VRAMQualitySource
16-bit34.9 GB38.2 GB39 GBthe original weightsComfy-Org/HunyuanImage_2.1_ComfyUI →
Q8_019.8 GB23.1 GB24 GBpractically identical to the originalcalcuis/hunyuanimage-gguf →
FP817.4 GB20.7 GB21 GBpractically identical to the originalComfy-Org/HunyuanImage_2.1_ComfyUI →
Q6_K15.7 GB19.0 GB19 GBvery close to the originalcalcuis/hunyuanimage-gguf →
Q5_K_M13.4 GB16.7 GB17 GBclose; small differences in fine detailcalcuis/hunyuanimage-gguf →
Q4_K_M11.3 GB14.6 GB15 GBgood; some loss in fine detail and textcalcuis/hunyuanimage-gguf →
Q3_K_M9.8 GB13.1 GB14 GBnoticeable loss of detailcalcuis/hunyuanimage-gguf →
Q2_K7.3 GB10.6 GB11 GBheavy loss; a last resortcalcuis/hunyuanimage-gguf →

“Needed” = file + 2.5 GB working memory + 0.8 GB system reserve.

02By amount of VRAM

6GBOffload onlyQ4_K_M8GBOffload onlyQ4_K_M10GBOffload onlyQ2_K12GBTightQ2_K16GBRunsQ4_K_M20GBRunsQ6_K24GBRuns wellQ8_032GBRuns wellQ8_0

03Best GPU for HunyuanImage 2.1

The cheapest cards (by launch price) that run it well, and every card sorted by memory: best GPU for HunyuanImage 2.1 → Planning bigger images or longer clips? Open the calculator →

04By graphics card

GPUVRAMVerdictBest fileNeeded
Desktop graphics cards
RTX 2060 6 GB6 GBOffload onlyQ4_K_M14.6 GB
RTX 3050 6 GB6 GBOffload onlyQ4_K_M14.6 GB
RTX 2070 Super 8 GB8 GBOffload onlyQ4_K_M14.6 GB
RTX 2080 Super 8 GB8 GBOffload onlyQ4_K_M14.6 GB
RTX 3050 8 GB8 GBOffload onlyQ4_K_M14.6 GB
RTX 3060 8 GB8 GBOffload onlyQ4_K_M14.6 GB
RTX 3060 Ti 8 GB8 GBOffload onlyQ4_K_M14.6 GB
RTX 3070 8 GB8 GBOffload onlyQ4_K_M14.6 GB
RTX 3070 Ti 8 GB8 GBOffload onlyQ4_K_M14.6 GB
RTX 4060 8 GB8 GBOffload onlyQ4_K_M14.6 GB
RTX 4060 Ti 8 GB8 GBOffload onlyQ4_K_M14.6 GB
RTX 5050 8 GB8 GBOffload onlyQ4_K_M14.6 GB
RTX 5060 8 GB8 GBOffload onlyQ4_K_M14.6 GB
RTX 5060 Ti 8 GB8 GBOffload onlyQ4_K_M14.6 GB
RX 7600 8 GB8 GBOffload onlyQ4_K_M14.6 GB
RX 9050 8 GB8 GBOffload onlyQ4_K_M14.6 GB
RX 9060 XT 8 GB8 GBOffload onlyQ4_K_M14.6 GB
Arc B570 10 GB10 GBOffload onlyQ2_K10.6 GB
RTX 3080 10 GB10 GBOffload onlyQ2_K10.6 GB
RTX 2080 Ti 11 GB11 GBTightQ2_K10.6 GB
Arc B580 12 GB12 GBTightQ2_K10.6 GB
RTX 2060 12 GB12 GBTightQ2_K10.6 GB
RTX 3060 12 GB12 GBTightQ2_K10.6 GB
RTX 3080 12 GB12 GBTightQ2_K10.6 GB
RTX 3080 Ti 12 GB12 GBTightQ2_K10.6 GB
RTX 4070 12 GB12 GBTightQ2_K10.6 GB
RTX 4070 Super 12 GB12 GBTightQ2_K10.6 GB
RTX 4070 Ti 12 GB12 GBTightQ2_K10.6 GB
RTX 5070 12 GB12 GBTightQ2_K10.6 GB
RX 7700 XT 12 GB12 GBTightQ2_K10.6 GB
RX 9070 GRE 12 GB12 GBTightQ2_K10.6 GB
Arc A770 16 GB16 GBRunsQ4_K_M14.6 GB
RTX 4060 Ti 16 GB16 GBRunsQ4_K_M14.6 GB
RTX 4070 Ti Super 16 GB16 GBRunsQ4_K_M14.6 GB
RTX 4080 16 GB16 GBRunsQ4_K_M14.6 GB
RTX 4080 Super 16 GB16 GBRunsQ4_K_M14.6 GB
RTX 5060 Ti 16 GB16 GBRunsQ4_K_M14.6 GB
RTX 5070 Ti 16 GB16 GBRunsQ4_K_M14.6 GB
RTX 5080 16 GB16 GBRunsQ4_K_M14.6 GB
RX 7600 XT 16 GB16 GBRunsQ4_K_M14.6 GB
RX 7800 XT 16 GB16 GBRunsQ4_K_M14.6 GB
RX 7900 GRE 16 GB16 GBRunsQ4_K_M14.6 GB
RX 9060 XT 16 GB16 GBRunsQ4_K_M14.6 GB
RX 9070 16 GB16 GBRunsQ4_K_M14.6 GB
RX 9070 XT 16 GB16 GBRunsQ4_K_M14.6 GB
RX 7900 XT 20 GB20 GBRunsQ6_K19.0 GB
Arc Pro B60 24 GB24 GBRuns wellQ8_023.1 GB
RTX 3090 24 GB24 GBRuns wellQ8_023.1 GB
RTX 3090 Ti 24 GB24 GBRuns wellQ8_023.1 GB
RTX 4090 24 GB24 GBRuns wellFP820.7 GB
RX 7900 XTX 24 GB24 GBRuns wellQ8_023.1 GB
Arc Pro B70 32 GB32 GBRuns wellQ8_023.1 GB
RTX 5090 32 GB32 GBRuns wellFP820.7 GB
Laptop GPUs
RTX 3050 Laptop 4 GB4 GBNot practicalQ2_K10.6 GB
RTX 3050 Ti Laptop 4 GB4 GBNot practicalQ2_K10.6 GB
RTX 2060 Laptop 6 GB6 GBOffload onlyQ4_K_M14.6 GB
RTX 3050 Laptop 6 GB6 GBOffload onlyQ4_K_M14.6 GB
RTX 3060 Laptop 6 GB6 GBOffload onlyQ4_K_M14.6 GB
RTX 4050 Laptop 6 GB6 GBOffload onlyQ4_K_M14.6 GB
RTX 2070 Laptop 8 GB8 GBOffload onlyQ4_K_M14.6 GB
RTX 2070 Super Laptop 8 GB8 GBOffload onlyQ4_K_M14.6 GB
RTX 2080 Laptop 8 GB8 GBOffload onlyQ4_K_M14.6 GB
RTX 2080 Super Laptop 8 GB8 GBOffload onlyQ4_K_M14.6 GB
RTX 3070 Laptop 8 GB8 GBOffload onlyQ4_K_M14.6 GB
RTX 3070 Ti Laptop 8 GB8 GBOffload onlyQ4_K_M14.6 GB
RTX 3080 Laptop 8 GB8 GBOffload onlyQ4_K_M14.6 GB
RTX 4060 Laptop 8 GB8 GBOffload onlyQ4_K_M14.6 GB
RTX 4070 Laptop 8 GB8 GBOffload onlyQ4_K_M14.6 GB
RTX 5050 Laptop 8 GB8 GBOffload onlyQ4_K_M14.6 GB
RTX 5060 Laptop 8 GB8 GBOffload onlyQ4_K_M14.6 GB
RTX 5070 Laptop 8 GB8 GBOffload onlyQ4_K_M14.6 GB
RX 7600M 8 GB8 GBOffload onlyQ4_K_M14.6 GB
RX 7600M XT 8 GB8 GBOffload onlyQ4_K_M14.6 GB
RX 7600S 8 GB8 GBOffload onlyQ4_K_M14.6 GB
RX 7700S 8 GB8 GBOffload onlyQ4_K_M14.6 GB
RTX 4080 Laptop 12 GB12 GBTightQ2_K10.6 GB
RTX 5070 Laptop 12 GB12 GBTightQ2_K10.6 GB
RTX 5070 Ti Laptop 12 GB12 GBTightQ2_K10.6 GB
RX 7800M 12 GB12 GBTightQ2_K10.6 GB
RTX 3080 Laptop 16 GB16 GBRunsQ4_K_M14.6 GB
RTX 3080 Ti Laptop 16 GB16 GBRunsQ4_K_M14.6 GB
RTX 4090 Laptop 16 GB16 GBRunsQ4_K_M14.6 GB
RTX 5080 Laptop 16 GB16 GBRunsQ4_K_M14.6 GB
RX 7900M 16 GB16 GBRunsQ4_K_M14.6 GB
RTX 5090 Laptop 24 GB24 GBRuns wellFP820.7 GB
Unified memory
Radeon 8060S (Strix Halo) 96 GB96 GBRuns well16-bit38.2 GB
Radeon 8065S (Gorgon Halo) 160 GB160 GBRuns well16-bit38.2 GB
Apple Silicon Macs (by memory)
Mac 16 GB12.7 GBTightQ2_K10.6 GB
Mac 18 GB14.4 GBTightQ3_K_M13.1 GB
Mac 24 GB19.6 GBRunsQ6_K19.0 GB
Mac 32 GB26.8 GBRuns wellQ8_023.1 GB
Mac 36 GB30.2 GBRuns wellQ8_023.1 GB
Mac 48 GB40.2 GBRuns well16-bit38.2 GB
Mac 64 GB55.7 GBRuns well16-bit38.2 GB
Mac 96 GB85 GBRuns well16-bit38.2 GB
Mac 128 GB115.4 GBRuns well16-bit38.2 GB
Mac 192 GB175.4 GBRuns well16-bit38.2 GB
Mac 256 GB236.9 GBRuns well16-bit38.2 GB
Mac 512 GB498.1 GBRuns well16-bit38.2 GB

05Text encoder, VAE and other files

FileFolderSizeWhen
Qwen2.5-VL 7B FP8
qwen_2.5_vl_7b_fp8_scaled.safetensors
models/text_encoders9.4 GBtext encoder · smaller, recommendedDownload →
Qwen2.5-VL 7B BF16
qwen_2.5_vl_7b.safetensors
models/text_encoders16.6 GBtext encoder · alternativeDownload →
ByT5 small Glyph-XL
byt5_small_glyphxl_fp16.safetensors
models/text_encoders0.4 GBrequiredDownload →
HunyuanImage 2.1 VAE
hunyuan_image_2.1_vae_fp16.safetensors
models/vae0.8 GBrequiredDownload →

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

06Where 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

07Measured and reported results

No measured results yet. Send yours.

08Training a LoRA for HunyuanImage 2.1

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

09Every file tracked for HunyuanImage 2.1

FileTypeSizeRepo
hunyuanimage2.1_distilled_bf16.safetensors (distilled)BF1634.91 GBComfy-Org/HunyuanImage_2.1_ComfyUI →
hunyuanimage2.1_bf16.safetensors (base)BF1634.85 GBComfy-Org/HunyuanImage_2.1_ComfyUI →
hunyuanimage2.1_refiner_bf16.safetensors (refiner)BF1630.08 GBComfy-Org/HunyuanImage_2.1_ComfyUI →
hunyuanimage2.1_distilled_fp8_e4m3fn.safetensors (distilled)FP817.45 GBComfy-Org/HunyuanImage_2.1_ComfyUI →
hunyuanimage2.1_fp8_e4m3fn.safetensors (base)FP817.43 GBComfy-Org/HunyuanImage_2.1_ComfyUI →
hunyuanimage2.1_refiner_fp8_e4m3fn.safetensors (refiner)FP815.04 GBComfy-Org/HunyuanImage_2.1_ComfyUI →
hunyuanimage2.1_fp32-f16.ggufF1635.90 GBcalcuis/hunyuanimage-gguf →
hunyuanimage2.1-q8_0.ggufQ8_019.79 GBcalcuis/hunyuanimage-gguf →
hunyuanimage2.1-q6_k.ggufQ6_K15.68 GBcalcuis/hunyuanimage-gguf →
hunyuanimage2.1-q5_1.ggufQ5_114.48 GBcalcuis/hunyuanimage-gguf →
hunyuanimage2.1-q5_0.ggufQ5_013.42 GBcalcuis/hunyuanimage-gguf →
hunyuanimage2.1-q5_k_m.ggufQ5_K_M13.42 GBcalcuis/hunyuanimage-gguf →
hunyuanimage2.1-q5_k_s.ggufQ5_K_S12.93 GBcalcuis/hunyuanimage-gguf →
hunyuanimage2.1-q4_1.ggufQ4_112.36 GBcalcuis/hunyuanimage-gguf →
hunyuanimage2.1-q4_0.ggufQ4_011.39 GBcalcuis/hunyuanimage-gguf →
hunyuanimage2.1-q4_k_m.ggufQ4_K_M11.30 GBcalcuis/hunyuanimage-gguf →
hunyuanimage2.1-q4_k_s.ggufQ4_K_S10.78 GBcalcuis/hunyuanimage-gguf →
hunyuanimage2.1-q3_k_m.ggufQ3_K_M9.75 GBcalcuis/hunyuanimage-gguf →
hunyuanimage2.1-q3_k_s.ggufQ3_K_S9.05 GBcalcuis/hunyuanimage-gguf →
hunyuanimage2.1-q2_k.ggufQ2_K7.32 GBcalcuis/hunyuanimage-gguf →

2K-native 17B DiT; optional refiner model. Text encoders: Qwen2.5-VL-7B + byT5 glyph encoder. HunyuanImage 3.0 (80B MoE) exists but is not a ComfyUI GGUF target.