the AI bench
VERIFIED JUNE 2026
All models

MODEL · HIDREAM · 17B (SAME ACROSS ALL THREE; DEV AND FAST ARE STEP-DISTILLED, NOT PRUNED)

HiDream-I1 (Full / Dev / Fast)

17B MMDiT open-weights image foundation, MIT-licensed, SOTA-at-release. Full/Dev/Fast ladder distilled down the step count. The MIT license is a big unlock for production commercial pipelines vs FLUX dev.

License: MIT (clean commercial — rare for 17B image model) · Context: Full ~50 steps / Dev ~28 / Fast ~16 · Released: April 7, 2025

The decision in five lines

The call
Skip for local — for image
Best for
image
Runs on
16 hardware picks fit (cheapest: Minisforum UM890 Pro · $463)
Watch out
At 12 GB VRAM — even Fast FP8 reports 16–18 GB.
Evidence
Estimated · last verified April 2026

17B (same across all three; Dev and Fast are step-distilled, not pruned)
PARAMETERS
IMAGE GEN
TYPE
Full
CONTEXT
~16–18 GB (FP8) / ~20–27 GB (BF16 Full)
VRAM AT Q4

Where we recommend this

Every tier slot in the planner where this model is a top or alternate pick. Pulled live from planner.js — when the planner refreshes, this table stays current.

IMAGE ·
HiDream-I1 Full (17B, FP16, MIT)April 2025 release; community testing shows it outperforms SDXL, DALL-E 3, and FLUX.1 on key benchmarks. MIT-licensed — clean for commercial redistribution. Frontier hardware unlocks the full FP16 weights.

The call

17B MMDiT open-weights image foundation, MIT-licensed, SOTA-at-release. Full/Dev/Fast ladder distilled down the step count. The MIT license is a big unlock for production commercial pipelines vs FLUX dev.

When not to use: At 12 GB VRAM — even Fast FP8 reports 16–18 GB. For those budgets, FLUX.2 klein 4B or Z-Image-Turbo are the sane picks. FLUX.2 [dev] still edges it on complex compositions.

Runner notes

ComfyUI. FP8 variants land at ~16–18 GB VRAM; BF16 full wants 20–27 GB. Fast is 1.5–2× quicker than Full at a small quality cost.

License
MIT (clean commercial — rare for 17B image model)
Released
April 7, 2025
Maker
HiDream

Hardware that fits

Every hardware pick whose memory fits this model at the quant we recommend. Sorted cheapest-first — the top row is your best-value fit. Click through for the full buyer’s guide.

Next step

Find-by-model — see what hardware runs this