Google has introduced Imagen 4, its latest text-to-image model, as a paid preview in the Gemini API, with limited free testing available in Google AI Studio starting June 24, 2025.
Imagen 4 can generate up to four 1024 脳 1024 images per call and accepts prompts up to 480 tokens. The pricing is approximately $0.04 per image, while the higher-fidelity version, Imagen 4 Ultra, costs about $0.06 per image.
Both versions include SynthID watermarks and offer near real-time generation. Imagen 4 Ultra provides tighter prompt alignment and optional 2K exports through Vertex AI.
Developers can access the same /generate endpoint used by Gemini models and may request higher rate limits during the preview period. The default limit is set at 20 requests per minute per project.
Early testers on social media platform X are already sharing side-by-side comparisons, highlighting sharper lettering and fewer artifacts compared to Imagen 3.
This launch enhances Google DeepMind鈥檚 creative stack, which now includes Imagen, Veo, and the Gemini model family. It is supported by generous free tiers and straightforward pay-as-you-go pricing on Google Cloud, aiming to attract both developers and enterprises.
