The Gemini 4 Carbon era of images, with models you can use today.

Gemini 4 Carbon is a reasoning, coding and agent model. Google’s announcement focuses on cybersecurity, long-horizon tasks and better usability for developers. It is not presented as a dedicated image model, and Gemini 4 Carbon is not publicly available yet, so no website can honestly offer “Gemini 4 Carbon image generation” today.

What early testers did share is telling. Community posts attributed to Gemini 4 Carbon show it writing code that draws things: SVG illustrations, three.js scenes and interactive diagrams. That is the Gemini 4 Carbon story for visuals so far: a model that plans and builds, rather than one that paints pixels directly.

Pixel-level image generation is handled by specialist models, and that is exactly what this page runs. If you want Google’s visual technology specifically, Nano Banana 2.1 and Nano Banana Pro are both available here. They are separate Google image models, not Gemini 4 Carbon, and we label them that way.

In short: read about Gemini 4 Carbon on our guide page, and make images here with models built for pixels. When Gemini 4 Carbon is open to everyone, Gemini 4 Carbon may become a great prompt planner for these image models.

Which model should you start with while Gemini 4 Carbon is unavailable?

• Nano Banana 2.1 for fast drafts and reference-based edits at 4 credits.
• GPT Image 2.5 when the prompt has many details that must all appear.
• Seedream 5.0 Pro and Qwen Image 3.0 for posters and legible text.
• FLUX 3 for photoreal products and controlled layouts.
• Midjourney for a painterly, cinematic look with four variations per run.

When Gemini 4 Carbon opens to the public through official Google channels, we will test whether Gemini 4 Carbon adds anything useful to an image workflow, such as prompt planning or layout code, and update this page. We won’t claim Gemini 4 Carbon support before then.

The habits that work with today’s image models will likely matter even more with a planning-heavy model like Gemini 4 Carbon: say what you want, in the order it matters. Clear, ordered prompts are also what Gemini 4 Carbon testers recommend for its code-drawn visuals.

1. Subject first. “A matte black ceramic cup” before “on warm linen”.
2. Then light and lens. “Soft daylight, 50mm, shallow depth of field.”
3. Name the medium. Photo, ink wash, isometric illustration, vector poster.
4. Quote exact text. Put words that must appear in quotes, and keep them short.
5. Use references for consistency. Upload a product or character image to keep it on-model.
6. Plan before you render. Outline the scene in a few lines first, the way Gemini 4 Carbon plans long tasks.

Need motion next? Take your still into the AI video generator as a first frame. Curious about the bigger picture? Read what Gemini 4 Carbon is.