whatmodel.sh / image / HiDream-I1-Full
HiDream-I1-Full
↗ share
+ contribute
HiDream-I1-Full is an open-weights image model from HiDream, released April 2025. Its published size is 17B parameters. Licence: mit.
Model information sources
Release date ↗ · Weights licence ↗ · Parameter count ↗
Stated for HiDream-I1 (sparse diffusion transformer) on the I1-Full card; excludes the Llama-3.1-8B / T5-XXL / CLIP text encoders and VAE.
Score within the LMArena text-to-image comparison set only.
$ how it compares
The image models ranked either side of HiDream-I1-Full, out of 197 with a score, then more from HiDream.
| # | model | score | where | ||
|---|---|---|---|---|---|
| 114 |  FLUX.1 Krea [dev] | 47±3 | api $0.025/image +local | ||
| 115 |  MetaQuery-XL | 47±6 | comparison only | ||
| 116 | HiDream-I1-Full | 47±2 | api $0.050/image +local | ||
| 117 |  GLIDE (filtered) | 47±15 | comparison only | ||
| 118 |  Ideogram 2.0 | 46±2 | api $0.080/image | ||
| more from HiDream | |||||
| 60 | HiDream-O1-Image | 58±2 | api $0.010/image +local | ||
| 86 | HiDream-O1-Image-Dev | 54±5 | comparison only | ||
| 190 | HiDream-E1.1 | 26±11 | comparison only | ||
$ what it's good at
text to image▂▅▅▃▄▄█▃▁▂▁▁ #76 of 126 · 83% of its score · 6 tests
Beats Latent Diffusion text2img-large (LDM-KL-8-G, LAION-400M), VQ-Diffusion (ITHQ) here, though they rank higher overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| GenExam (relaxed overall, GPT-5 judge) | 21.2% ✎ | ▄█▂▃▂▂▆▄·▂·▂ beats 7/16 | 43% |
| T2I-ReasonBench (overall) | 64.2% ✎ | ▅▂█▂▃▂····▂▂ beats 10/13 | 23% |
| MMMG knowledge-image generation (MMMG-Score, average over education levels) | 28.0% ✎ | ▃··▃·█▃▃▅··▃ beats 9/9 | 8.9% |
| DPG-Bench lab | 85.9% ✎ | ▄·▂·▂▆█▄▂▄▂▂ beats 8/17 | 4.8% |
| LongText-Bench (EN) lab | 54.3% ✎ | ▃·····▅██▅▃▃ beats 0/12 | 2.7% |
| LongText-Bench (CN) lab | 2.4% ✎ | ▃·······████ beats 0/12 | 0.3% |
other tests1 not in a task · 16% of its score
| test | result | vs the others in its run | counts |
|---|---|---|---|
| VVRBench accuracy (10,000 verifiable synthetic-scene tasks) | 4.2% ✎ | ▃·█▅▃·▅▃···▃ beats 5/10 | 16% |
$ where to run it
Through an API
api · fal.ai · $0.050/image · fal.ai ↗ · price sourceUSD for one 1-megapixel output: fal bills this endpoint $0.05 per megapixel (price = rate × 1 MP; fal's rounding for 1024×1024 ≈ 1.05 MP is not stated; larger outputs cost proportionally more). Product: fal.ai serverless text-to-image. API model: fal-ai/hidream-i1-full (HiDream I1 Full, 17B; the dev and fast endpoints are separate). Settings: 1 MP output, one image. Price basis: published per-megapixel list rate × 1 MP. Currency: USD. Not a HiDream direct tariff. fal.ai official model page, observed 2026-09-26.
Other ways to run it
- Run it as published · needs ~76.4 GB, NVIDIA only · from huggingface.co/HiDream-ai
git clone https://github.com/HiDream-ai/HiDream-I1 && cd HiDream-I1 && pip install -r requirements.txt && python ./inference.py --model_type full
Weights and documentation: huggingface.co/HiDream-ai ↗
$ how the score was worked out · GenExam: A Multidisciplinary… 43%, T2I-ReasonBench: Benchmarking… 23%, Verifiable Visual Rewards Tra… 16% and 2 more
Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works
| source | results | share of the score |
|---|---|---|
 GenExam: A Multidisciplinary Text-to-Image Exam (arXiv v5) | 1 | 43% |
 T2I-ReasonBench: Benchmarking Reasoning-Informed Text-to-Image Generation (Findings of ACL 2026) | 1 | 23% |
| Verifiable Visual Rewards Transfer from Synthetic Scenes to Natural Prompts (VVRBench), arXiv:2609.35641v1 | 1 | 16% |
| MMMG: A Massive, Multidisciplinary, Multi-Tier Generation Benchmark for Text-to-Image Reasoning (arXiv:2506.10963v2) | 1 | 9% |
 HiDream-ai/HiDream-O1-Image repository README results · lab-reported | 3 | 8% |
| prior · pull toward the middle | 1% | |
| weighted average | = 46.6 |
$ improve this page
Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.
data updated · sources and licences · json