The split that matters most in image generation: cloud tools you use in a browser, versus open-weight models you can download and run on your own graphics card. Open-weight describes the license, not the only way to use a model, because most open models are also offered as paid cloud APIs.
As of the snapshot date, the leading closed models are GPT Image 2 (inside ChatGPT), Midjourney V8, and Google Nano Banana Pro. The leading open-weight models are FLUX.2, Qwen-Image, Ideogram 4, and Z-Image, all runnable on consumer or prosumer GPUs.
Free image generation is easy to find. Ideogram gives 10 credits a day, Recraft 30, Leonardo 150 tokens, and ChatGPT and Gemini both generate images on their free tiers. Midjourney is the notable exception, with no free tier at all.
Snapshot of 7 June 2026. 22 entries in this group, about 19 with a free option and 7 that can run on your own hardware. Prices and versions change often, and each name links to its official page. The full directory has search and filters across all categories.
How to Choose
Just make images now:
ChatGPT (GPT Image 2) and Gemini generate on their free tiers with nothing to install.
Text inside the image:
Ideogram, Qwen-Image, and Recraft specialize in clean lettering and typography. Recraft also outputs vector SVG.
Commercial safety:
Adobe Firefly is trained on licensed and Adobe Stock data and is positioned for commercial use, with Creative Cloud integration.
Run it on your own GPU:
FLUX.2 klein runs on a gaming GPU with about 13 GB of VRAM, Z-Image on about 16 GB, and Stable Diffusion 3.5 from about 8 GB. No per-image fee after the hardware.
Highest aesthetic polish:
Midjourney V8, subscription only, from 10 dollars a month.
All 22 Image generation Entries
| Tool | Maker | What it is | Cost | Where it runs | Hardware |
|---|---|---|---|---|---|
| Adobe Firefly | Adobe | Trained on licensed data for commercial safety. Creative Cloud integration. | Free + paid | Cloud | Any device |
| Canva (Magic Media) | Canva | Image generation built into the Canva design editor. | Free + paid | Cloud | Any device |
| FLUX.2 [dev] | Black Forest Labs | Open 32B model for generation and single or multi-reference editing. | Open-weight | Cloud or local | High-end GPU (RTX 4090/5090) |
| FLUX.2 [klein] | Black Forest Labs | Fastest FLUX, sub-second generation on consumer hardware. | Open-weight | Cloud or local | Gaming GPU ~13GB |
| FLUX.2 [pro] | Black Forest Labs | Top-tier photorealism and prompt adherence. The hosted commercial tier. | Paid | Cloud | Any device |
| GPT Image 2 | OpenAI | Reasoning-based generation with near-perfect text rendering and 4K output. | Free + paid | Cloud | Any device |
| Grok Imagine | xAI | Image and video generation inside Grok and X. | Free + paid | Cloud | Any device |
| HunyuanImage 3.0 | Tencent | Largest open text-to-image model. Not runnable on a single consumer GPU. | Open-weight | Local | Data-center GPUs (~3x80GB) |
| Ideogram 4 | Ideogram | Best-in-class in-image text and typography. Now open-weight. | Open-weight | Cloud or local | High-end GPU |
| Imagen 4 | Strong photorealism and text rendering, via the Gemini API. | Free + paid | Cloud | Any device | |
| Krea | Krea AI | Aggregator of 150+ models plus a real-time canvas. | Free + paid | Cloud | Any device |
| Leonardo.Ai | Leonardo (Canva) | Game and asset-oriented platform with its own models. | Free + paid | Cloud | Any device |
| Magnific (Freepik) | Freepik | Aggregator running its own Mystic model plus 30+ third-party models. | Free + paid | Cloud | Any device |
| Midjourney | Midjourney | Strongest aesthetic quality and atmosphere. Discord and web app. | Paid | Cloud | Any device |
| Nano Banana (Gemini Image) | Conversational image generation and editing in Gemini. Pro adds high-res. | Free + paid | Cloud | Any device | |
| Playground | Playground | Design-focused editor routing to several top models. | Free + paid | Cloud | Any device |
| Qwen-Image | Alibaba | Open model with excellent complex text rendering, including Chinese. | Open-weight | Cloud or local | High-end GPU (20B) |
| Recraft | Recraft | Raster plus native vector and SVG output, with brand style sets. | Free + paid | Cloud | Any device |
| Reve | Reve AI | Strong prompt adherence and typography from a Palo Alto startup. | Free + paid | Cloud | Any device |
| Seedream | ByteDance | High-resolution (up to 4K) generation and editing. | Paid | Cloud | Any device |
| Stable Diffusion 3.5 | Stability AI | The largest open LoRA and fine-tune ecosystem. Runs locally. | Open-weight | Cloud or local | Gaming GPU 8GB+ |
| Z-Image | Alibaba | Efficient 6B open model with top open photorealism. Runs on consumer cards. | Open-weight | Cloud or local | Gaming GPU ~16GB |
Common Questions
Can I generate AI images for free?
Yes. ChatGPT and Gemini generate on free tiers, Ideogram gives 10 credits a day, Recraft 30 a day (public, non-commercial), and Leonardo 150 tokens a day. Free tiers usually publish your images publicly or restrict commercial use, so read the plan terms.
What do I need to run image models locally?
A gaming GPU. Stable Diffusion 3.5 runs from about 8 GB of VRAM, FLUX.2 klein needs about 13 GB, and Z-Image about 16 GB. Once running, there is no per-image cost.
Can I use the images commercially?
It depends on the tool and the plan. Several free tiers exclude commercial use (Recraft's free plan, for example), while Adobe Firefly is explicitly positioned as commercially safe. Check the license of the specific tool and tier before publishing.
Keep Reading
Verified articles related to this page:
Other model pages: