Habr compares four image generators on Russian prompts
A Habr article reports a 14-task comparison of Nano Banana 2.1, Nano Banana 2, Nano Banana Pro and GPT Image 2.5 Sunburst, run through SYNTX.AI on October 12, 2026. Tasks covered counting, Russian text in posters and infographics, diagrams and tables, image edits, and a three-turn character dialogue. Eight tasks placed Russian text directly in the image.
The authors scored 76 binary checklist items. GPT Image 2.5 Sunburst scored 72, Nano Banana 2 scored 71, and Nano Banana 2.1 and Nano Banana Pro each scored 67. They say the 1–5 point spread, with one generation per task and one evaluator, does not establish a winner. Errors varied by task: Pro failed sharply on the 18-element table, while 2.1 lost capitalization and punctuation in poster and infographic outputs. The article says generation time was not measured and the comparison was not blind.