The open-weight design model that renders posters, logos, packaging and multi-line text cleanly. Three speed tiers, five aspect ratios, no watermark. From 1 credit per image on Aura AI, on every plan.
Ideogram V4, or Ideogram 4.0, is the fourth generation of the image model from Ideogram, the company that made readable text in AI images its specialty. It was released on June 3, 2026, and it is the largest architectural change in the company's history: rather than iterating on Ideogram 3, the team trained a new 9.3 billion parameter foundation model from scratch. It is a fully single-stream diffusion transformer with 34 layers that processes text and image tokens as one sequence, and it uses Qwen3-VL-8B-Instruct as its text encoder, reading features from 13 intermediate layers. The weights are public on GitHub and Hugging Face under a non-commercial license, with commercial licensing available from Ideogram.
The design focus shows in what the model is good at. Ideogram describes best-in-class in-image text generation: signage, logos, captions, watermarks and multi-line copy that reads clean on posters, packaging and brand work. In an independent evaluation by ContraLabs, professional designers picked Ideogram 4 first for typography work 47.9 percent of the time against competing models, and on DesignArena it ranks as the top open-weight text-to-image model. It also understands structured JSON prompts, so you can specify a color palette, a bounding-box layout and the typographic treatment in one request. Upstream it renders any resolution from 256 to 2048 pixels at multiples of 16, with aspect ratios up to 6:1, and its best quality is at 2048 x 2048.
On Aura AI, Ideogram V4 runs in the text-to-image tool in three tiers: the full Ideogram V4 at 2 credits per image, and V4 Instant and V4 Fast at 1 credit each for quicker drafts. All three offer five aspect ratios and deliver output without watermark. The same credits work on Nano Banana, Recraft, GPT Image 2 and every other model in the studio, so you can draft a layout with Ideogram and animate it with Seedance or Kling without leaving the platform.
Six capabilities that separate 4.0 from Ideogram 3 and from general-purpose image models
Multi-line headlines, small captions, logos and signage render legibly and stay in the right place. Naming a typeface style in the prompt works better than a generic description like "bold sans-serif".
V4 was trained on structured captions, so it accepts JSON prompts that set composition, lighting, color palette, typography and bounding-box positions for each element in a single request.
Hand the model your brand colors and the output stays on palette. Combined with layout control, this is what makes it usable for real brand systems rather than one-off art.
Ideogram exposes fast, balanced and quality rendering upstream. On Aura AI that maps to V4 Instant and V4 Fast at 1 credit for drafts and the full V4 at 2 credits for the final render.
Upstream the model handles any size from 256 to 2048 pixels in multiples of 16, with aspect ratios up to 6:1. Ultra-wide banners and tall posters are native shapes, not crops.
The 9.3B model is downloadable in nf4 and fp8 builds and works with Diffusers. Teams can fine-tune it on their own style guides and product photography, which is rare among top-ranked image models.
Four steps from a design brief to a finished poster, logo or ad
Flat cost per image, no resolution or aspect ratio multiplier
| Variant | Mode | Credits per image | Best for |
|---|---|---|---|
| Ideogram V4 | Text to image | 2 credits | Final posters, logos, packaging, dense typography |
| Ideogram V4 Instant | Text to image | 1 credit | Fastest drafts, thumbnails, exploring headline options |
| Ideogram V4 Fast | Text to image | 1 credit | Quick social posts and layout iterations |
How the new Ideogram release compares to its predecessor and to other design-capable models on Aura AI
| Feature | Ideogram V4 | Ideogram 3 | Nano Banana | Recraft V4 |
|---|---|---|---|---|
| Maker | Ideogram | Ideogram | Google DeepMind | Recraft |
| Latest release | Jun 2026 | 2025 | Feb 2026 (Nano Banana 2) | May 2026 (V4.1) |
| Credits per image | 2 (Instant, Fast: 1) | 1 | 3 | 2 (V4.1), 3 (V4) |
| Speed tiers on Aura AI | 3 | 1 | 2 | 2 |
| Open weights | ✓ Yes, 9.3B | No | No | No |
| Typography focus | ✓ Core strength | ✓ Core strength | ✓ Multilingual | ✓ Design, vector |
| Structured JSON prompts | ✓ Layout, palette, type | No | No | No |
| Plan | All plans | All plans | All plans | All plans |
Where typography, layout control and palette conditioning pay off
Event posters with a headline, a date, a venue and a call to action, all readable in one generation. Ideogram V4 keeps the hierarchy: the big line stays big, the small print stays small and legible.
Explore dozens of logo directions with V4 Instant at 1 credit, then render the shortlist with the full V4. Wordmarks come out with correct letters, which is where most image models still stumble.
Product labels need a name, a variant, a weight and a tagline in the right spots. Bounding-box layout control lets you place each element instead of hoping the model guesses.
Quote cards, promo tiles and story ads with on-brand colors. Palette conditioning keeps every variation on brand, and V4 Fast makes a full week of posts cheap to produce.
Common questions about Ideogram V4 on Aura AI
Posters, logos and packaging with text that reads. From 1 credit per image, no watermark, on every plan, worldwide.