Synthesis history

5 versions, newest first.

Version 5 2026-07-06 02:22 UTC · 158 items

New items this pass include a TechRadar hands-on comparison of full Nano Banana 2 against ChatGPT Images 2.0 [^39844], a K-Dense benchmark targeting scientific image generation with Nano Banana 2 Lite [^39843], and addi…
Version 4 2026-07-04 18:46 UTC · 152 items

The new items this pass are almost entirely comparison content — blog posts, YouTube videos, an OpenRouter comparison page, and a Reddit thread — pitting Nano Banana 2 Lite and the full Nano Banana 2 against GPT Image 2…
Version 3 2026-07-03 08:38 UTC · 142 items

The new items this pass are largely amplification rather than substantive new claims: fal.ai now hosts Nano Banana 2 Lite as an API endpoint [^38923], Google published an official Gemini Omni Flash model card [^38319], …
Version 2 2026-07-02 02:39 UTC · 129 items

The most significant new development is the Gemini Omni Flash prompt-rejection bug: users reported at launch that most simple video editing prompts were flagged as policy violations [^37115][^37118], Google confirmed it…
Version 1 2026-06-30 18:21 UTC · 84 items

Google DeepMind launched two generative media models on June 30, 2026: Nano Banana 2 Lite, a text-to-image model generating outputs in under 4 seconds at $0.034 per 1,000 images [^36666][^36716], and Gemini Omni Flash, …