On July 1, 2026, Google DeepMind released the image generation model 'Nano Banana 2 Lite' and the video generation model 'Gemini Omni Flash.' Nano Banana 2 Lite, officially named 'Gemini 3.1 Flash-Lite Image,' is based on Gemini 3.1 Flash-Lite, handles contexts of up to 1 million tokens, and outputs both images and text with a maximum of 4,000 tokens for image output and 64,000 tokens for text output. It can generate images from text in approximately 4 seconds, faster than Nano Banana 2 (approximately 20 seconds) and the original Nano Banana (approximately 7 seconds). In Arena.ai Elo scores, Nano Banana 2 Lite scored 1251 for image generation, 19 points short of Nano Banana 2's 1270 but 100 points higher than the original's 1151; in image editing, it scored 1308, 13 points higher than the original's 1295. The cost of outputting a single 1K resolution image is just over 3 cents for Nano Banana 2 Lite, compared to approximately 3.9 cents for Nano Banana 2 and approximately 6.7 cents for the original Nano Banana. Images generated by Nano Banana 2 Lite are embedded with an invisible digital watermark called 'SynthID.'