Ideogram

We gave four frontier image editing models the same photo and 30 edits in a row: Ideogram 4.5 keeps most of the room intact, while the others drift, GPT Image 2.5 Sunburst most visibly.

We've seen some interesting demos of multi-turn editing consistency from the latest image editing models, so we ran our own test: GPT Image 2.5 Sunburst, #1 on our Image Editing leaderboard, against Ideogram 4.5, FLUX 3 and Nano Banana 2.1. Our leaderboard scores single edits; this tests what happens when 30 consecutive changes stack up, with each model editing its own previous output through a 30-step real estate staging sequence: light the fire, add a sofa, repaint the walls, swap day for twilight, and more.

Why are the final results so different?

@ideogram_ai's Ideogram 4.5 and @bfl_ml's FLUX 3 edit locally. On small edits like adding a vase of tulips, we measured that they left 95% or more of the image essentially untouched. GPT Image 2.5 (Sunburst) re-renders most of the entire image on every edit, leaving only about a fifth of the image unchanged, so small shifts in colour and detail compound over turns. Nano Banana 2.1 sits in between: its edits stay local, but the rest of the image shifts slightly and gradually darkens.

You can use Ideogram 4.5 to transform famous paintings into photos using this one prompt:

"Turn this painting into a realistic scene."

1/ The Scream, Edvard Munch, 1893 🧵

I spent thousands of hours photographing and editing as a hobby. It brought out my creativity and attention to detail.

With so much code in my head, I want to return to photography to relax. But I have hundreds of unedited photos because I can't justify the editing time.

Really hope Ideogram works as advertised. Open weights are coming, and I'm looking forward to hopefully running it locally to EDIT my photos, not GENERATE them.

BREAKING: Ideogram 4.5 by @ideogram_ai places 15th in Image Editing with an Elo of 1250.

This puts the model in the same performance band as MAI-Image-2.6 by @MicrosoftAI and Gemini 3 Pro Image Preview by @GoogleDeepMind. The model performs particularly well with typography edits, such as modifying text in infographics.

Congratulations to the @ideogram_ai team on the launch!

It was a pleasure hosting Canada's AI Minister @EvanLSolomon at Ideogram's Toronto HQ. We talked about the 3 C's and 2 T's Canadian AI needs: capital, compute, customers, talent, and trust. Proud of our superstar team, building sovereign generative media here at home. 🇨🇦

Enterprise-grade AI visual design, powered by AMD Instinct.

Ideogram just released Ideogram 4.0, a 9.3B-parameter open-weight Diffusion Transformer that outperforms all open-weight image models in LMArena and Design Arena, with a 0.97 X-Omni English OCR text-rendering score (vs. an industry average of 0.30–0.50).

The engine behind it? AMD Instinct MI350X GPUs, 288 GB of HBM3 memory and 8 TB/s of bandwidth, delivering the throughput enterprise diffusion workloads demand.

Read how @ideogram_ai and AMD are scaling production-grade visual AI from millions to billions of images per month 👉 bit.ly/3TosU1m

The best way to bring the composition from your head into an image → Ideogram V4 + drawing bounding boxes in Comfy.

The control here is quite unique. The model uses structured JSON so drawing bounding boxes to get the exact placement works very well.

The model only needs 12 steps (turbo), so iterating with different seeds + very impressive text rendering capability leads me to say this is a state of the art open source image model right now.

Using the 'Ideogram 4 Prompt Builder' node by Kijai.

Image generation is becoming serious enterprise infrastructure.

We asked @mo_norouzi, CEO of @Ideogram_ai, about the inflection point:

"Until last year it was all about beauty, entertainment, maybe a small business uses it for a landing page. But it's not possible to do on-brand imagery at the quality the best brands want."

"Now the quality is at the level that enterprise is super excited to use this."

"It's not about what's the best generic model, it's about what is the best model we can build for you, for your enterprise, for your use case."

"We're gonna have a lot of branded models. Creatives will build their own models, design becomes even more important than ever. Customization is the new frontier."

Hey @Google, What's is the best open weight image model?