Datasheet for local AI49 models98 GPUsData read 2026-09-25
No adsNo tracking
Check my GPU →

Illustrious XL / Pony (SDXL anime) VRAM requirements

The two big anime families built on SDXL. Almost every anime checkpoint on Civitai (WAI, NoobAI and friends) has this same size and fits exactly like SDXL.

Released 2024-09Licence: Illustrious: Fair AI Public License 1.0-SD; Pony V6: original Civitai license (HF mirror metadata says cdla-permissive-2.0, unofficial mirror)Steps: 25–30Text encoder: built into the checkpoint
TypeImage
Parameters3.5B
Fits entirely from3 GB
8-bit or better from5 GB

01The files, and how much VRAM each needs

FileSizeNeededMin. VRAMQualitySource
16-bit6.9 GB7.1 GB8 GBthe original weightsOnomaAIResearch/Illustrious-xl-early-release-v0 →
Q8_02.7 GB4.7 GB5 GBpractically identical to the originalcalcuis/illustrious →
FP83.5 GB5.5 GB6 GBpractically identical to the originalcalcuis/illustrious →
Q6_K2.1 GB4.1 GB5 GBvery close to the originalcalcuis/illustrious →
Q5_K_M1.8 GB3.8 GB4 GBclose; small differences in fine detailcalcuis/illustrious →
Q4_K_M1.5 GB3.5 GB4 GBgood; some loss in fine detail and textcalcuis/illustrious →
Q3_K_M1.2 GB3.2 GB4 GBnoticeable loss of detailcalcuis/illustrious →
Q2_K0.9 GB2.9 GB3 GBheavy loss; a last resortcalcuis/illustrious →

“Needed” = file + 1.2 GB working memory + 0.8 GB system reserve. The 6.94 GB checkpoint holds the UNet plus text encoders and VAE; only the UNet (about 5.1 GB) sits in VRAM while sampling. The GGUF and FP8 files from calcuis contain the UNet only.

03Best GPU for Illustrious / Pony

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

04By graphics card

GPUVRAMVerdictBest fileNeeded
Desktop graphics cards
RTX 2060 6 GB6 GBRuns wellQ8_04.7 GB
RTX 3050 6 GB6 GBRuns wellQ8_04.7 GB
RTX 2070 Super 8 GB8 GBRuns well16-bit7.1 GB
RTX 2080 Super 8 GB8 GBRuns well16-bit7.1 GB
RTX 3050 8 GB8 GBRuns well16-bit7.1 GB
RTX 3060 8 GB8 GBRuns well16-bit7.1 GB
RTX 3060 Ti 8 GB8 GBRuns well16-bit7.1 GB
RTX 3070 8 GB8 GBRuns well16-bit7.1 GB
RTX 3070 Ti 8 GB8 GBRuns well16-bit7.1 GB
RTX 4060 8 GB8 GBRuns well16-bit7.1 GB
RTX 4060 Ti 8 GB8 GBRuns well16-bit7.1 GB
RTX 5050 8 GB8 GBRuns well16-bit7.1 GB
RTX 5060 8 GB8 GBRuns well16-bit7.1 GB
RTX 5060 Ti 8 GB8 GBRuns well16-bit7.1 GB
RX 7600 8 GB8 GBRuns well16-bit7.1 GB
RX 9050 8 GB8 GBRuns well16-bit7.1 GB
RX 9060 XT 8 GB8 GBRuns well16-bit7.1 GB
Arc B570 10 GB10 GBRuns well16-bit7.1 GB
RTX 3080 10 GB10 GBRuns well16-bit7.1 GB
RTX 2080 Ti 11 GB11 GBRuns well16-bit7.1 GB
Arc B580 12 GB12 GBRuns well16-bit7.1 GB
RTX 2060 12 GB12 GBRuns well16-bit7.1 GB
RTX 3060 12 GB12 GBRuns well16-bit7.1 GB
RTX 3080 12 GB12 GBRuns well16-bit7.1 GB
RTX 3080 Ti 12 GB12 GBRuns well16-bit7.1 GB
RTX 4070 12 GB12 GBRuns well16-bit7.1 GB
RTX 4070 Super 12 GB12 GBRuns well16-bit7.1 GB
RTX 4070 Ti 12 GB12 GBRuns well16-bit7.1 GB
RTX 5070 12 GB12 GBRuns well16-bit7.1 GB
RX 7700 XT 12 GB12 GBRuns well16-bit7.1 GB
RX 9070 GRE 12 GB12 GBRuns well16-bit7.1 GB
Arc A770 16 GB16 GBRuns well16-bit7.1 GB
RTX 4060 Ti 16 GB16 GBRuns well16-bit7.1 GB
RTX 4070 Ti Super 16 GB16 GBRuns well16-bit7.1 GB
RTX 4080 16 GB16 GBRuns well16-bit7.1 GB
RTX 4080 Super 16 GB16 GBRuns well16-bit7.1 GB
RTX 5060 Ti 16 GB16 GBRuns well16-bit7.1 GB
RTX 5070 Ti 16 GB16 GBRuns well16-bit7.1 GB
RTX 5080 16 GB16 GBRuns well16-bit7.1 GB
RX 7600 XT 16 GB16 GBRuns well16-bit7.1 GB
RX 7800 XT 16 GB16 GBRuns well16-bit7.1 GB
RX 7900 GRE 16 GB16 GBRuns well16-bit7.1 GB
RX 9060 XT 16 GB16 GBRuns well16-bit7.1 GB
RX 9070 16 GB16 GBRuns well16-bit7.1 GB
RX 9070 XT 16 GB16 GBRuns well16-bit7.1 GB
RX 7900 XT 20 GB20 GBRuns well16-bit7.1 GB
Arc Pro B60 24 GB24 GBRuns well16-bit7.1 GB
RTX 3090 24 GB24 GBRuns well16-bit7.1 GB
RTX 3090 Ti 24 GB24 GBRuns well16-bit7.1 GB
RTX 4090 24 GB24 GBRuns well16-bit7.1 GB
RX 7900 XTX 24 GB24 GBRuns well16-bit7.1 GB
Arc Pro B70 32 GB32 GBRuns well16-bit7.1 GB
RTX 5090 32 GB32 GBRuns well16-bit7.1 GB
Laptop GPUs
RTX 3050 Laptop 4 GB4 GBRunsQ5_K_M3.8 GB
RTX 3050 Ti Laptop 4 GB4 GBRunsQ5_K_M3.8 GB
RTX 2060 Laptop 6 GB6 GBRuns wellQ8_04.7 GB
RTX 3050 Laptop 6 GB6 GBRuns wellQ8_04.7 GB
RTX 3060 Laptop 6 GB6 GBRuns wellQ8_04.7 GB
RTX 4050 Laptop 6 GB6 GBRuns wellFP85.5 GB
RTX 2070 Laptop 8 GB8 GBRuns well16-bit7.1 GB
RTX 2070 Super Laptop 8 GB8 GBRuns well16-bit7.1 GB
RTX 2080 Laptop 8 GB8 GBRuns well16-bit7.1 GB
RTX 2080 Super Laptop 8 GB8 GBRuns well16-bit7.1 GB
RTX 3070 Laptop 8 GB8 GBRuns well16-bit7.1 GB
RTX 3070 Ti Laptop 8 GB8 GBRuns well16-bit7.1 GB
RTX 3080 Laptop 8 GB8 GBRuns well16-bit7.1 GB
RTX 4060 Laptop 8 GB8 GBRuns well16-bit7.1 GB
RTX 4070 Laptop 8 GB8 GBRuns well16-bit7.1 GB
RTX 5050 Laptop 8 GB8 GBRuns well16-bit7.1 GB
RTX 5060 Laptop 8 GB8 GBRuns well16-bit7.1 GB
RTX 5070 Laptop 8 GB8 GBRuns well16-bit7.1 GB
RX 7600M 8 GB8 GBRuns well16-bit7.1 GB
RX 7600M XT 8 GB8 GBRuns well16-bit7.1 GB
RX 7600S 8 GB8 GBRuns well16-bit7.1 GB
RX 7700S 8 GB8 GBRuns well16-bit7.1 GB
RTX 4080 Laptop 12 GB12 GBRuns well16-bit7.1 GB
RTX 5070 Laptop 12 GB12 GBRuns well16-bit7.1 GB
RTX 5070 Ti Laptop 12 GB12 GBRuns well16-bit7.1 GB
RX 7800M 12 GB12 GBRuns well16-bit7.1 GB
RTX 3080 Laptop 16 GB16 GBRuns well16-bit7.1 GB
RTX 3080 Ti Laptop 16 GB16 GBRuns well16-bit7.1 GB
RTX 4090 Laptop 16 GB16 GBRuns well16-bit7.1 GB
RTX 5080 Laptop 16 GB16 GBRuns well16-bit7.1 GB
RX 7900M 16 GB16 GBRuns well16-bit7.1 GB
RTX 5090 Laptop 24 GB24 GBRuns well16-bit7.1 GB
Unified memory
Radeon 8060S (Strix Halo) 96 GB96 GBRuns well16-bit7.1 GB
Apple Silicon Macs (by memory)
Mac 16 GB12.7 GBRuns well16-bit7.1 GB
Mac 18 GB14.4 GBRuns well16-bit7.1 GB
Mac 24 GB19.6 GBRuns well16-bit7.1 GB
Mac 32 GB26.8 GBRuns well16-bit7.1 GB
Mac 36 GB30.2 GBRuns well16-bit7.1 GB
Mac 48 GB40.2 GBRuns well16-bit7.1 GB
Mac 64 GB55.7 GBRuns well16-bit7.1 GB
Mac 96 GB85 GBRuns well16-bit7.1 GB
Mac 128 GB115.4 GBRuns well16-bit7.1 GB
Mac 192 GB175.4 GBRuns well16-bit7.1 GB
Mac 256 GB236.9 GBRuns well16-bit7.1 GB
Mac 512 GB498.1 GBRuns well16-bit7.1 GB

On RTX 40/50 GPUs the FP8 file is preferred over Q8_0 when both fit (hardware FP8). All verdicts are calculated; see how the numbers work.

05Text encoder, VAE and other files

The text encoder is inside the checkpoint, so there is nothing extra to download.

06Where the files go in ComfyUI

FileFolderLoader node
The checkpoint (.safetensors)ComfyUI/models/checkpointsLoad Checkpoint
GGUF file (.gguf)ComfyUI/models/unetUnet Loader (GGUF) — from the ComfyUI-GGUF node pack

Standard ComfyUI folders. After copying files, press R in ComfyUI (or restart it) to refresh the lists. Some uploads need their uploader's own loader node — see the notes above.

07AMD, Intel and NVIDIA: which file types are fast

File typeRTX 50RTX 40RTX 30 / 20RX 9000RX 7000/6000 · Strix HaloIntel Arc
16-bitRunsRunsRunsRunsRunsRuns
FP8Native FP8Native FP8No FP8 speed-upNative FP8No FP8 speed-upNo FP8 speed-up
GGUFRuns (GGUF node)Runs (GGUF node)Runs (GGUF node)Runs (GGUF node)Runs (GGUF node)Runs (GGUF node)

Every file type loads on every listed GPU, so the memory verdicts apply to all of them. What differs is speed: FP8 maths needs RTX 40/50 or RX 9000 (with ROCm 6.4+ and PyTorch 2.7+); ComfyUI's INT8 maths runs on NVIDIA and AMD, not on Intel; GGUF is unpacked on the fly on any GPU, which costs some speed. NVFP4 files are fast only on RTX 50. AMD runs ComfyUI on Windows through ROCm, Intel through PyTorch XPU; some custom nodes are NVIDIA-only. Source: ComfyUI model_management.py. AMD and Intel guide →

08Measured and reported results

LabelGPUSetupResultPeak VRAMDateSource
reportedRTX 3060 Ti 8 GBIllustrious-XL-v2.0 · 1024x1024
“NVIDIA RTX 3060 Ti | 1.96it/s | 0.51s/it | CUDA 12.9 | ComfyUI (Unknown) | Ubuntu Server 24.04.2 LTS”
Community speed list; ComfyUI, Euler/Normal, CFG 8, square 1:1, batch speed only (no per-image time); date = list last-updated date; SDXL 1024px section
0.51 s/it—2025-11-29huggingface.co →
reportedRTX 3090 24 GBIllustrious-XL-v2.0 · 1024x1024
“NVIDIA RTX 3090 | 4.00it/s | 0.25s/it | CUDA 12.9 | ComfyUI (Unknown) | Arch Linux”
Community speed list; ComfyUI, Euler/Normal, CFG 8, square 1:1, batch speed only (no per-image time); date = list last-updated date; SDXL 1024px section
0.25 s/it—2025-11-29huggingface.co →
reportedRTX 4060 Laptop 8 GBWAI-Illustrious SDXL v16.0 · 1024x1024 · 20 steps
“1024x1024 | 1.47 | 13s | 15.81s | ~5.6GB”
ComfyUI, euler_ancestral/Karras CFG 5; columns it/s | KSampler | total | VRAM; no --lowvram; VRAM approximate
15.81 s / image · 1.47 it/s5.6 GB2026-02-26lilting.ch →
reportedRTX 4090 24 GBIllustrious-XL-v2.0 · 1024x1024
“NVIDIA RTX 4090 | 7.00it/s | 0.14s/it | CUDA 12.9 | ComfyUI (Unknown) | Windows 11 24H2”
Community speed list; ComfyUI, Euler/Normal, CFG 8, square 1:1, batch speed only (no per-image time); date = list last-updated date; SDXL 1024px section
0.14 s/it—2025-11-29huggingface.co →
reportedRTX 5060 Ti 16 GBIllustrious-XL-v2.0 · 1024x1024
“NVIDIA RTX 5060 Ti | 2.60it/s | 0.39s/it | CUDA 12.9 | ComfyUI (Unknown) | Arch Linux”
Community speed list; ComfyUI, Euler/Normal, CFG 8, square 1:1, batch speed only (no per-image time); date = list last-updated date; SDXL 1024px section; row says 'RTX 5060 Ti', co
0.39 s/it—2025-11-29huggingface.co →
reportedRTX 5090 32 GBIllustrious-XL-v2.0 · 1024x1024
“NVIDIA RTX 5090 | 8.95it/s | 0.11s/it | CUDA 12.8 | ComfyUI (Unknown) | Windows 11 24H2”
Community speed list; ComfyUI, Euler/Normal, CFG 8, square 1:1, batch speed only (no per-image time); date = list last-updated date; SDXL 1024px section
0.11 s/it—2025-11-29huggingface.co →

Reported results are other people's numbers, copied as published, with a link. Settings, drivers and ComfyUI versions differ, so compare them with care. Send yours.

09Training a LoRA for Illustrious / Pony

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

10Every file tracked for Illustrious / Pony

FileTypeSizeRepo
ponyDiffusionV6XL_v6StartWithThisOne.safetensors (pony-v6)16-bit6.94 GBLyliaEngine/Pony_Diffusion_V6_XL →
all-in-one checkpoint: includes text encoder(s) and VAE, not diffusion-model-only
Illustrious-XL-v0.1.safetensors (illustrious-v0.1)16-bit6.94 GBOnomaAIResearch/Illustrious-xl-early-release-v0 →
all-in-one checkpoint: includes text encoder(s) and VAE, not diffusion-model-only (CLIP-L + CLIP-G + VAE)
Illustrious-XL-v0.1-GUIDED.safetensors (illustrious-v0.1-guided)16-bit6.94 GBOnomaAIResearch/Illustrious-xl-early-release-v0 →
all-in-one checkpoint: includes text encoder(s) and VAE, not diffusion-model-only
illustrious-f16.gguf (illustrious)F165.14 GBcalcuis/illustrious →
illustrious_fp8_e4m3fn.safetensors (illustrious)FP83.47 GBcalcuis/illustrious →
all-in-one checkpoint: includes text encoder(s) and VAE, not diffusion-model-only (per repo README: fp8 file includes VAE and CLIPs)
illustrious-q8_0.gguf (illustrious)Q8_02.74 GBcalcuis/illustrious →
illustrious-q6_k.gguf (illustrious)Q6_K2.12 GBcalcuis/illustrious →
illustrious-q5_1.gguf (illustrious)Q5_11.94 GBcalcuis/illustrious →
illustrious-q5_k_m.gguf (illustrious)Q5_K_M1.81 GBcalcuis/illustrious →
illustrious-q5_0.gguf (illustrious)Q5_01.78 GBcalcuis/illustrious →
illustrious-q5_k_s.gguf (illustrious)Q5_K_S1.78 GBcalcuis/illustrious →
illustrious-q4_1.gguf (illustrious)Q4_11.62 GBcalcuis/illustrious →
illustrious-q4_k_m.gguf (illustrious)Q4_K_M1.53 GBcalcuis/illustrious →
illustrious-q4_k_s.gguf (illustrious)Q4_K_S1.46 GBcalcuis/illustrious →
illustrious-q4_0.gguf (illustrious)Q4_01.46 GBcalcuis/illustrious →
illustrious-q3_k_l.gguf (illustrious)Q3_K_L1.19 GBcalcuis/illustrious →
illustrious-q3_k_m.gguf (illustrious)Q3_K_M1.15 GBcalcuis/illustrious →
illustrious-q3_k_s.gguf (illustrious)Q3_K_S1.12 GBcalcuis/illustrious →
illustrious-q2_k.gguf (illustrious)Q2_K0.89 GBcalcuis/illustrious →

SDXL architecture (UNet ~2.6B, 3.5B incl. text encoders) - fits exactly like SDXL. Illustrious XL and Pony V6 are the two main anime base families; most people run Civitai fine-tunes (WAI, NoobAI, etc.) that are the same 6.94 GB all-in-one size. HF downloads: Pony mirror ~289k, Illustrious v0.1 ~81k.