Krea 2 (Turbo) VRAM requirements
Krea's own 12B model from June 2026, one of the most downloaded ComfyUI releases of the year. Turbo is the 8-step distilled version; Raw is the base for fine-tuning, same size.
TypeImage
Parameters12.8B
Fits entirely from8 GB
8-bit or better from16 GB
01Measured on my RTX 5060 Ti 16 GB
INT8 · 1024×1024FASTEST
10.9s / image
krea2_turbo_int8_convrot.safetensors
Peak VRAM15.3 GB
Run range10.9–11.0 s
FP8 · 1024×1024
17.7s / image
krea2_turbo_fp8_scaled.safetensors
Peak VRAM14.9 GB
Run range17.7–17.8 s
02The files, and how much VRAM each needs
| File | Size | Needed | Min. VRAM | Quality | Source |
|---|---|---|---|---|---|
| 16-bit | 26.3 GB | 28.9 GB | 29 GB | the original weights | Comfy-Org/Krea-2 → |
| Q8_0 | 13.7 GB | 16.3 GB | 17 GB | practically identical to the original | vantagewithai/Krea-2-Turbo-GGUF → |
| FP8 | 13.1 GB | 15.7 GB | 16 GB | practically identical to the original | Comfy-Org/Krea-2 → |
| INT8 | 13.5 GB | 16.1 GB | 17 GB | practically identical to the original | Comfy-Org/Krea-2 → |
| Q6_K | 10.6 GB | 13.2 GB | 14 GB | very close to the original | vantagewithai/Krea-2-Turbo-GGUF → |
| Q5_K_M | 8.9 GB | 11.5 GB | 12 GB | close; small differences in fine detail | vantagewithai/Krea-2-Turbo-GGUF → |
| Q4_K_M | 7.5 GB | 10.1 GB | 11 GB | good; some loss in fine detail and text | vantagewithai/Krea-2-Turbo-GGUF → |
| Q3_K_M | 6.0 GB | 8.6 GB | 9 GB | noticeable loss of detail | vantagewithai/Krea-2-Turbo-GGUF → |
| Q2_K | 4.9 GB | 7.5 GB | 8 GB | heavy loss; a last resort | vantagewithai/Krea-2-Turbo-GGUF → |
“Needed” = file + 1.8 GB working memory + 0.8 GB system reserve.
Free ComfyUI workflow, tested on my RTX 5060 Ti: INT8 workflow (.json). Drag the file into ComfyUI. All workflows →
03By amount of VRAM
04Best GPU for Krea 2
The cheapest cards (by launch price) that run it well, and every card sorted by memory: best GPU for Krea 2 → Planning bigger images or longer clips? Open the calculator →
05By graphics card
| GPU | VRAM | Verdict | Best file | Needed |
|---|---|---|---|---|
| Desktop graphics cards | ||||
| RTX 2060 6 GB | 6 GB | Offload only | Q2_K | 7.5 GB |
| RTX 3050 6 GB | 6 GB | Offload only | Q2_K | 7.5 GB |
| RTX 2070 Super 8 GB | 8 GB | Tight | Q2_K | 7.5 GB |
| RTX 2080 Super 8 GB | 8 GB | Tight | Q2_K | 7.5 GB |
| RTX 3050 8 GB | 8 GB | Tight | Q2_K | 7.5 GB |
| RTX 3060 8 GB | 8 GB | Tight | Q2_K | 7.5 GB |
| RTX 3060 Ti 8 GB | 8 GB | Tight | Q2_K | 7.5 GB |
| RTX 3070 8 GB | 8 GB | Tight | Q2_K | 7.5 GB |
| RTX 3070 Ti 8 GB | 8 GB | Tight | Q2_K | 7.5 GB |
| RTX 4060 8 GB | 8 GB | Tight | Q2_K | 7.5 GB |
| RTX 4060 Ti 8 GB | 8 GB | Tight | Q2_K | 7.5 GB |
| RTX 5050 8 GB | 8 GB | Tight | Q2_K | 7.5 GB |
| RTX 5060 8 GB | 8 GB | Tight | Q2_K | 7.5 GB |
| RTX 5060 Ti 8 GB | 8 GB | Tight | Q2_K | 7.5 GB |
| RX 7600 8 GB | 8 GB | Tight | Q2_K | 7.5 GB |
| RX 9050 8 GB | 8 GB | Tight | Q2_K | 7.5 GB |
| RX 9060 XT 8 GB | 8 GB | Tight | Q2_K | 7.5 GB |
| Arc B570 10 GB | 10 GB | Tight | Q3_K_M | 8.6 GB |
| RTX 3080 10 GB | 10 GB | Tight | Q3_K_M | 8.6 GB |
| RTX 2080 Ti 11 GB | 11 GB | Runs | Q4_K_M | 10.1 GB |
| Arc B580 12 GB | 12 GB | Runs | Q5_K_M | 11.5 GB |
| RTX 2060 12 GB | 12 GB | Runs | Q5_K_M | 11.5 GB |
| RTX 3060 12 GB | 12 GB | Runs | Q5_K_M | 11.5 GB |
| RTX 3080 12 GB | 12 GB | Runs | Q5_K_M | 11.5 GB |
| RTX 3080 Ti 12 GB | 12 GB | Runs | Q5_K_M | 11.5 GB |
| RTX 4070 12 GB | 12 GB | Runs | Q5_K_M | 11.5 GB |
| RTX 4070 Super 12 GB | 12 GB | Runs | Q5_K_M | 11.5 GB |
| RTX 4070 Ti 12 GB | 12 GB | Runs | Q5_K_M | 11.5 GB |
| RTX 5070 12 GB | 12 GB | Runs | Q5_K_M | 11.5 GB |
| RX 7700 XT 12 GB | 12 GB | Runs | Q5_K_M | 11.5 GB |
| RX 9070 GRE 12 GB | 12 GB | Runs | Q5_K_M | 11.5 GB |
| Arc A770 16 GB | 16 GB | Runs well | FP8 | 15.7 GB |
| RTX 4060 Ti 16 GB | 16 GB | Runs well | FP8 | 15.7 GB |
| RTX 4070 Ti Super 16 GB | 16 GB | Runs well | FP8 | 15.7 GB |
| RTX 4080 16 GB | 16 GB | Runs well | FP8 | 15.7 GB |
| RTX 4080 Super 16 GB | 16 GB | Runs well | FP8 | 15.7 GB |
| RTX 5060 Ti 16 GB | 16 GB | Runs well | FP8 | 15.7 GB |
| RTX 5070 Ti 16 GB | 16 GB | Runs well | FP8 | 15.7 GB |
| RTX 5080 16 GB | 16 GB | Runs well | FP8 | 15.7 GB |
| RX 7600 XT 16 GB | 16 GB | Runs well | FP8 | 15.7 GB |
| RX 7800 XT 16 GB | 16 GB | Runs well | FP8 | 15.7 GB |
| RX 7900 GRE 16 GB | 16 GB | Runs well | FP8 | 15.7 GB |
| RX 9060 XT 16 GB | 16 GB | Runs well | FP8 | 15.7 GB |
| RX 9070 16 GB | 16 GB | Runs well | FP8 | 15.7 GB |
| RX 9070 XT 16 GB | 16 GB | Runs well | FP8 | 15.7 GB |
| RX 7900 XT 20 GB | 20 GB | Runs well | Q8_0 | 16.3 GB |
| Arc Pro B60 24 GB | 24 GB | Runs well | Q8_0 | 16.3 GB |
| RTX 3090 24 GB | 24 GB | Runs well | Q8_0 | 16.3 GB |
| RTX 3090 Ti 24 GB | 24 GB | Runs well | Q8_0 | 16.3 GB |
| RTX 4090 24 GB | 24 GB | Runs well | FP8 | 15.7 GB |
| RX 7900 XTX 24 GB | 24 GB | Runs well | Q8_0 | 16.3 GB |
| Arc Pro B70 32 GB | 32 GB | Runs well | 16-bit | 28.9 GB |
| RTX 5090 32 GB | 32 GB | Runs well | 16-bit | 28.9 GB |
| Laptop GPUs | ||||
| RTX 3050 Laptop 4 GB | 4 GB | Offload only | Q4_K_M | 10.1 GB |
| RTX 3050 Ti Laptop 4 GB | 4 GB | Offload only | Q4_K_M | 10.1 GB |
| RTX 2060 Laptop 6 GB | 6 GB | Offload only | Q2_K | 7.5 GB |
| RTX 3050 Laptop 6 GB | 6 GB | Offload only | Q2_K | 7.5 GB |
| RTX 3060 Laptop 6 GB | 6 GB | Offload only | Q2_K | 7.5 GB |
| RTX 4050 Laptop 6 GB | 6 GB | Offload only | Q2_K | 7.5 GB |
| RTX 2070 Laptop 8 GB | 8 GB | Tight | Q2_K | 7.5 GB |
| RTX 2070 Super Laptop 8 GB | 8 GB | Tight | Q2_K | 7.5 GB |
| RTX 2080 Laptop 8 GB | 8 GB | Tight | Q2_K | 7.5 GB |
| RTX 2080 Super Laptop 8 GB | 8 GB | Tight | Q2_K | 7.5 GB |
| RTX 3070 Laptop 8 GB | 8 GB | Tight | Q2_K | 7.5 GB |
| RTX 3070 Ti Laptop 8 GB | 8 GB | Tight | Q2_K | 7.5 GB |
| RTX 3080 Laptop 8 GB | 8 GB | Tight | Q2_K | 7.5 GB |
| RTX 4060 Laptop 8 GB | 8 GB | Tight | Q2_K | 7.5 GB |
| RTX 4070 Laptop 8 GB | 8 GB | Tight | Q2_K | 7.5 GB |
| RTX 5050 Laptop 8 GB | 8 GB | Tight | Q2_K | 7.5 GB |
| RTX 5060 Laptop 8 GB | 8 GB | Tight | Q2_K | 7.5 GB |
| RTX 5070 Laptop 8 GB | 8 GB | Tight | Q2_K | 7.5 GB |
| RX 7600M 8 GB | 8 GB | Tight | Q2_K | 7.5 GB |
| RX 7600M XT 8 GB | 8 GB | Tight | Q2_K | 7.5 GB |
| RX 7600S 8 GB | 8 GB | Tight | Q2_K | 7.5 GB |
| RX 7700S 8 GB | 8 GB | Tight | Q2_K | 7.5 GB |
| RTX 4080 Laptop 12 GB | 12 GB | Runs | Q5_K_M | 11.5 GB |
| RTX 5070 Laptop 12 GB | 12 GB | Runs | Q5_K_M | 11.5 GB |
| RTX 5070 Ti Laptop 12 GB | 12 GB | Runs | Q5_K_M | 11.5 GB |
| RX 7800M 12 GB | 12 GB | Runs | Q5_K_M | 11.5 GB |
| RTX 3080 Laptop 16 GB | 16 GB | Runs well | FP8 | 15.7 GB |
| RTX 3080 Ti Laptop 16 GB | 16 GB | Runs well | FP8 | 15.7 GB |
| RTX 4090 Laptop 16 GB | 16 GB | Runs well | FP8 | 15.7 GB |
| RTX 5080 Laptop 16 GB | 16 GB | Runs well | FP8 | 15.7 GB |
| RX 7900M 16 GB | 16 GB | Runs well | FP8 | 15.7 GB |
| RTX 5090 Laptop 24 GB | 24 GB | Runs well | FP8 | 15.7 GB |
| Unified memory | ||||
| Radeon 8060S (Strix Halo) 96 GB | 96 GB | Runs well | 16-bit | 28.9 GB |
| Radeon 8065S (Gorgon Halo) 160 GB | 160 GB | Runs well | 16-bit | 28.9 GB |
| Apple Silicon Macs (by memory) | ||||
| Mac 16 GB | 12.7 GB | Runs | Q5_K_M | 11.5 GB |
| Mac 18 GB | 14.4 GB | Runs | Q6_K | 13.2 GB |
| Mac 24 GB | 19.6 GB | Runs well | Q8_0 | 16.3 GB |
| Mac 32 GB | 26.8 GB | Runs well | Q8_0 | 16.3 GB |
| Mac 36 GB | 30.2 GB | Runs well | 16-bit | 28.9 GB |
| Mac 48 GB | 40.2 GB | Runs well | 16-bit | 28.9 GB |
| Mac 64 GB | 55.7 GB | Runs well | 16-bit | 28.9 GB |
| Mac 96 GB | 85 GB | Runs well | 16-bit | 28.9 GB |
| Mac 128 GB | 115.4 GB | Runs well | 16-bit | 28.9 GB |
| Mac 192 GB | 175.4 GB | Runs well | 16-bit | 28.9 GB |
| Mac 256 GB | 236.9 GB | Runs well | 16-bit | 28.9 GB |
| Mac 512 GB | 498.1 GB | Runs well | 16-bit | 28.9 GB |
06Text encoder, VAE and other files
| File | Folder | Size | When | |
|---|---|---|---|---|
| Qwen3-VL 4B FP8 qwen3vl_4b_fp8_scaled.safetensors | models/text_encoders | 5.2 GB | text encoder · smaller, recommended | Download → |
| Qwen3-VL 4B BF16 qwen3vl_4b_bf16.safetensors | models/text_encoders | 8.9 GB | text encoder · alternative | Download → |
| Qwen-Image VAE qwen_image_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).
07Where 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 |
08Measured and reported results
| Label | GPU | Setup | Result | Peak VRAM | Date | Source |
|---|---|---|---|---|---|---|
| measured | RTX 5060 Ti 16 GB | krea2_turbo_int8_convrot.safetensors · 1024x1024 · 8 steps 3 timed runs after 1 warm-up (15.56 s incl. loading). Text encoder Qwen3-VL 4B FP8 (qwen3vl_4b_fp8_scaled), Qwen-Image VAE. 8 steps, euler/simple, CFG 1. ComfyUI 0.38.1 portable, P | 10.95 s / image | 15.3 GB | 2026-10-04 | |
| measured | RTX 5060 Ti 16 GB | krea2_turbo_fp8_scaled.safetensors · 1024x1024 · 8 steps 3 timed runs after 1 warm-up (22.59 s incl. loading). Text encoder Qwen3-VL 4B FP8 (qwen3vl_4b_fp8_scaled), Qwen-Image VAE. 8 steps, euler/simple, CFG 1. ComfyUI 0.38.1 portable, P | 17.74 s / image | 14.9 GB | 2026-10-04 | |
| reported | RTX 3060 Ti 8 GB | fp8 “3060 ti here, 8GB VRAM, fp8 quant gets about 3.9 seconds per iteration” Posted on Krea-2-Turbo repo; ComfyUI ('run ... in Comfy now'); resolution/steps not stated; year inferred from Krea 2 release (Jun 2026) | 3.9 s/it | — | 2026-07-01 | huggingface.co → |
| reported | RTX 5070 Ti 16 GB | Krea 2 Turbo (ComfyUI release, 17.8GB total download) · 1024x1024 · 8 steps “generating a 1024x1024 pixel image took 16-25 seconds when the prompts were rewritten, and approximately 9-15 seconds when the same prompts were reused” GIGAZINE review, ComfyUI; range only (9-15 s same prompt, 16-25 s new prompt incl. text encoding) | 9–15 s / image | — | 2026-06-24 | gigazine.net → |
Other people's numbers are copied as published, with a link. Send yours.
09Training a LoRA for Krea 2
| Trainer | VRAM | Type | Settings and quote | Source |
|---|---|---|---|---|
| OneTrainer | 16 GB | example run | Official preset (lowest Krea 2 preset): 512px, batch 2, transformer INT_W8A8, TE fp8, layer offload fraction 0.3, torch.compile “#krea2 LoRA 16GB.json” | github.com → |
| SimpleTuner | 18.1 GB | example run | Measured on H100 80GB: int8-torchao, 512px, batch 1, fused QKV, gradient checkpointing, compile off (1024px batch 1: 20.35 GiB; bf16 512px: 31.10 GiB) “int8-torchao: 512: 1: 0.535: 18.10 GiB” | github.com → |
| SimpleTuner | 24 GB | stated minimum | Reduced resolution or quantised; 48GB+ recommended for comfortable 512px “the realistic minimum: an NVIDIA GPU with at least 24GB VRAM for reduced-resolution or quantised experiments” | github.com → |
| diffusion-pipe | 24 GB | example run | fp8 diffusion model, rank 32, 512px “This configuration can train a rank 32 LoRA at 512 resolution with 24GB VRAM.” | github.com → |
reported From each trainer's own docs, 2026-09-25. Why the figures differ →
10Every file tracked for Krea 2
| File | Type | Size | Repo |
|---|---|---|---|
| krea2_turbo_bf16.safetensors (turbo) | BF16 | 26.28 GB | Comfy-Org/Krea-2 → |
| krea2_raw_bf16.safetensors (raw) | BF16 | 26.28 GB | Comfy-Org/Krea-2 → |
| krea2_turbo_mxfp8.safetensors (turbo) | MXFP8 | 13.53 GB | Comfy-Org/Krea-2 → |
| krea2_turbo_int8_convrot.safetensors (turbo) | INT8 | 13.49 GB | Comfy-Org/Krea-2 → |
| krea2_raw_int8_convrot.safetensors (raw) | INT8 | 13.49 GB | Comfy-Org/Krea-2 → |
| krea2_turbo_fp8_scaled.safetensors (turbo) | FP8 | 13.14 GB | Comfy-Org/Krea-2 → |
| krea2_raw_fp8_scaled.safetensors (raw) | FP8 | 13.14 GB | Comfy-Org/Krea-2 → |
| krea2_turbo_nvfp4.safetensors (turbo) | NVFP4 | 7.67 GB | Comfy-Org/Krea-2 → |
| krea2_turbo-Q8_0.gguf (turbo) | Q8_0 | 13.71 GB | vantagewithai/Krea-2-Turbo-GGUF → |
| krea2_turbo-Q6_K.gguf (turbo) | Q6_K | 10.58 GB | vantagewithai/Krea-2-Turbo-GGUF → |
| krea2_turbo-Q5_1.gguf (turbo) | Q5_1 | 9.67 GB | vantagewithai/Krea-2-Turbo-GGUF → |
| krea2_turbo-Q5_0.gguf (turbo) | Q5_0 | 8.87 GB | vantagewithai/Krea-2-Turbo-GGUF → |
| krea2_turbo-Q5_K_M.gguf (turbo) | Q5_K_M | 8.87 GB | vantagewithai/Krea-2-Turbo-GGUF → |
| krea2_turbo-Q5_K_S.gguf (turbo) | Q5_K_S | 8.87 GB | vantagewithai/Krea-2-Turbo-GGUF → |
| krea2_turbo-Q4_1.gguf (turbo) | Q4_1 | 8.18 GB | vantagewithai/Krea-2-Turbo-GGUF → |
| krea2_turbo-Q4_0.gguf (turbo) | Q4_0 | 7.49 GB | vantagewithai/Krea-2-Turbo-GGUF → |
| krea2_turbo-Q4_K_M.gguf (turbo) | Q4_K_M | 7.49 GB | vantagewithai/Krea-2-Turbo-GGUF → |
| krea2_turbo-Q4_K_S.gguf (turbo) | Q4_K_S | 7.49 GB | vantagewithai/Krea-2-Turbo-GGUF → |
| krea2_turbo-Q3_K_M.gguf (turbo) | Q3_K_M | 6.01 GB | vantagewithai/Krea-2-Turbo-GGUF → |
| krea2_turbo-Q3_K_S.gguf (turbo) | Q3_K_S | 6.01 GB | vantagewithai/Krea-2-Turbo-GGUF → |
| krea2_turbo-Q2_K.gguf (turbo) | Q2_K | 4.89 GB | vantagewithai/Krea-2-Turbo-GGUF → |
12B DiT from Krea (HF API total 12.8B). Turbo = 8-step distilled; Raw = undistilled base for fine-tuning. VAE: qwen_image_vae.safetensors (Qwen-Image VAE). Native ComfyUI support; GGUF via ComfyUI-GGUF.