HiDream-I1 is an image generation model by HiDream.ai (17.1B; text encoder: Llama-3.1-8B + T5-XXL + CLIP). Uses four text encoders, so the encoders are large.
Change precision, resolution →
| Precision | Model size | 1024×1024 | 1536×1536 |
|---|---|---|---|
| BF16 / FP16 (uncompressed) | 34.2 GB | 35.5 GB all in VRAM 49.8 GB | 36.4 GB all in VRAM 50.7 GB |
| FP8 (half size, near-lossless) | 17.4 GB | 18.7 GB all in VRAM 33.1 GB | 19.6 GB all in VRAM 34.0 GB |
| GGUF Q8_0 (similar to FP8) | 18.3 GB | 19.6 GB all in VRAM 33.9 GB | 20.5 GB all in VRAM 34.8 GB |
| GGUF Q5_K_S | 12.1 GB | 14.4 GB all in VRAM 27.8 GB | 14.4 GB all in VRAM 28.7 GB |
| GGUF Q4_K_S (popular low-VRAM choice) | 10.3 GB | 14.4 GB all in VRAM 25.9 GB | 14.4 GB all in VRAM 26.8 GB |
Bold = VRAM needed when ComfyUI swaps the text encoder and the model. "All in VRAM" = keeping both loaded. Text encoder assumed at FP8.
| GPU | VRAM | Best precision | VRAM needed | Verdict | Buy |
|---|---|---|---|---|---|
| RTX 4060 Ti 16GB previous gen | 16 GB | GGUF Q5_K_S | 14.4 GB | ○ Runs with model swapping | Check price |
| Radeon RX 7600 XT previous gen | 16 GB | GGUF Q5_K_S | 14.4 GB | ○ Runs with model swapping | Check price |
| Radeon RX 9060 XT 16GB | 16 GB | GGUF Q5_K_S | 14.4 GB | ○ Runs with model swapping | Check price |
| RTX 5060 Ti 16GB | 16 GB | GGUF Q5_K_S | 14.4 GB | ○ Runs with model swapping | Check price |
| Radeon RX 6800 previous gen | 16 GB | GGUF Q5_K_S | 14.4 GB | ○ Runs with model swapping | Check price |
| Radeon RX 7800 XT previous gen | 16 GB | GGUF Q5_K_S | 14.4 GB | ○ Runs with model swapping | Check price |
| Radeon RX 9070 XT | 16 GB | GGUF Q5_K_S | 14.4 GB | ○ Runs with model swapping | Check price |
| Radeon RX 9070 | 16 GB | GGUF Q5_K_S | 14.4 GB | ○ Runs with model swapping | Check price |
| RTX 4070 Ti SUPER previous gen | 16 GB | GGUF Q5_K_S | 14.4 GB | ○ Runs with model swapping | Check price |
| RTX 4080 previous gen | 16 GB | GGUF Q5_K_S | 14.4 GB | ○ Runs with model swapping | Check price |
| RTX 4080 SUPER previous gen | 16 GB | GGUF Q5_K_S | 14.4 GB | ○ Runs with model swapping | Check price |
| RTX 5070 Ti | 16 GB | GGUF Q5_K_S | 14.4 GB | ○ Runs with model swapping | Check price |
| RTX 5080 | 16 GB | GGUF Q5_K_S | 14.4 GB | ○ Runs with model swapping | Check price |
| Radeon RX 7900 XT previous gen | 20 GB | FP8 | 18.7 GB | ○ Runs with model swapping | Check price |
| RTX 3090 previous gen | 24 GB | FP8 | 18.7 GB | ○ Runs with model swapping | Check price |
| Radeon RX 7900 XTX previous gen | 24 GB | FP8 | 18.7 GB | ○ Runs with model swapping | Check price |
| RTX 4090 previous gen | 24 GB | FP8 | 18.7 GB | ○ Runs with model swapping | Check price |
| RTX 5090 | 32 GB | FP8 | 18.7 GB | ○ Runs with model swapping | Check price |
ComfyUI also runs on AMD Radeon (ROCm) and Intel Arc, but setup is harder than on NVIDIA and some features (such as FP8 acceleration) are not supported.
→ Image & Video AI VRAM Calculator (all 15 models) / Local LLM VRAM Calculator
Estimates for ComfyUI, about ±20%. Model specs from the publisher's config files and released file sizes (checked October 2026).