HiDream-I1-Full (17B-A5.5B MoE) VRAM & GPU Requirement Calculator

Interactive hardware sizing for HiDream-I1-Full (17B-A5.5B MoE) (17B by HiDream AI). Estimated 8K context footprint: 11.4 GB (INT4) • 18.6 GB (FP8) • 35.5 GB (FP16).

GPUs

Memory Requirements 18.55 GB

Requires 1 GPUs (based on memory capacity)

17 GB

All model weights

0 GB

Conversation history cache

0.85 GB

Expert model optimization

0.7 GB

Temporary computation cache

ComfyUI Pipeline & Hardware Estimates
Est. Gen Time
4.8 s/img
Min System RAM
48 GB RAM
Resolution
1 MP
Offload Mode
Smart Offload
Quick Deploy CLI Command
python main.py --normalvram --fp8_e4m3fn-unet --fp8_e4m3fn-text-enc --preview-method auto
Flags reflect your selected precision, context length, and tensor parallel GPU count.Open HiDream-I1-Full spec page →

HiDream-I1-Full (17B-A5.5B MoE) Architecture & Memory Profile

Total Parameters
17B
Architecture
MoE (5.5B active)
Developer / Provider
HiDream AI
Recommended Precision
FP8

HiDream-I1-Full (17B-A5.5B MoE) ComfyUI VRAM Matrix by Precision & Resolution

Precomputed 3-stage ComfyUI Smart Offload + VAE Tiling VRAM peak (recommended host RAM: 48 GB).

Precision1024x1024 (1MP)1536x1536 (2.25MP)2048x2048 (4MP)LoRA Train (1MP)
FP1635.5 GB36.8 GB38.9 GB39.3 GB
FP818.6 GB19.8 GB21.9 GB22.3 GB
INT818.6 GB19.8 GB21.9 GB22.3 GB
INT411.4 GB12.7 GB14.8 GB15.1 GB

Recommended GPU Hardware for HiDream-I1-Full (17B-A5.5B MoE) (FP8)

Minimum GPU count for 1024x1024 generation with ComfyUI offload and estimated seconds/image.

GPU HardwareMin GPUsTotal VRAM PoolEst. Gen Time
NVIDIA RTX 4090 (24GB)1×24 GB4.8 s/img
NVIDIA RTX 5090 (32GB)1×32 GB3.5 s/img
Apple M4 Max (128GB) (128GB)1×128 GB5.9 s/img
NVIDIA A100 (80GB) (80GB)1×80 GB5.1 s/img
NVIDIA H100 (80GB)1×80 GB1.4 s/img
NVIDIA H200 (141GB)1×141 GB1.4 s/img
Huawei Ascend 910B (64GB)1×64 GB2.7 s/img

HiDream-I1-Full (17B-A5.5B MoE) Deployment FAQ

How much VRAM is needed to run HiDream-I1-Full (17B-A5.5B MoE)?
At 1024x1024 with ComfyUI Smart Offload and VAE Tiling enabled, HiDream-I1-Full (17B-A5.5B MoE) needs about 35.5 GB in BF16/FP16, 18.6 GB in FP8, and 11.4 GB in INT4/GGUF (paired with ~48 GB system RAM).
Can I run HiDream-I1-Full (17B-A5.5B MoE) on RTX 4090 or H100 GPUs?
Running HiDream-I1-Full (17B-A5.5B MoE) in FP8 precision requires at least 1x RTX 4090 (24GB) or 1x NVIDIA H100 (80GB) GPUs.
What precision should I choose for HiDream-I1-Full (17B-A5.5B MoE) in production?
For HiDream-I1-Full (17B-A5.5B MoE) (17B by HiDream AI), FP8 provides the best balance between inference speed, VRAM footprint, and output quality.