Imagen 4
2025-05-20 · Publisher : Google · Domain : Artificial Intelligence
Google DeepMind unveils Imagen 4, its next-generation image generation model: finer details, markedly improved text rendering and outputs up to 2K, offered in Standard, Ultra and Fast.
What changes
- High sharpness on details (fabrics, water droplets, fur) in photorealistic and abstract styles.
- Marked improvement in spelling and typography within images (maps, posters, comics).
- Several image formats and resolution up to 2K.
- June 24, 2025: paid preview in the Gemini API with Imagen 4 and Imagen 4 Ultra (more faithful to prompts).
- August 14-15, 2025: general availability and launch of Imagen 4 Fast, which is faster.
### Specifications - Availability at I/O: Gemini app, Whisk, Vertex AI, Google Slides, Vids, Docs. - Gemini API and AI Studio since June 2025. - Outputs marked with the SynthID watermark. - Google announces Imagen 4 Fast up to 10 times faster than Imagen 3. ### Pricing | Model | USD per generated image | |---|---| | Imagen 4 | 0.04 | | Imagen 4 Ultra | 0.06 | | Imagen 4 Fast | 0.02 | ### Benchmarks No numerical score published. ### Key takeaways An image model suited to content with legible text, at a fixed price per image, until Nano Banana (Gemini 2.5 Flash Image) takes over for editing.
Links
Tags : Image, Google, DeepMind, Imagen, API