發布日期2026/02
排名表現
評測分析
Nano Banana 2 is a strong option for image processing workflows, offering competitive performance in both image editing and generation. Best used standalone for vision tasks or paired with other models to cover non-vision domains.
強項
- Outstanding image editing capabilities, ranking high on official leaderboards
- Strong image generation, competing directly with leading models
- Fast, well-integrated vision capabilities suitable for real-time tasks
弱項
- Lacks standout benchmark rankings outside image processing (complex reasoning, code generation, and other specialized tasks)
- Context length may be limited compared to flagship models, impacting large-scale project workflows
- Output quality in complex edge cases still falls short of high-end models
適用情境
Fast image editing and processing: inpainting, retouching, and removing unwanted objectsGenerating images for rapid prototyping, ideation, and time-sensitive content creationImage content analysis, recognition, and information extraction
指南與影片
Nano Banana 2 (Gemini 3.1 Flash Image) is Google's AI image model, accessible via Google AI Studio (aistudio.google.com) or the Gemini API. You can enter text prompts to generate high-quality images, edit existing images with simple instructions, and generate images from video. The model supports four resolutions (512p, 1K, 2K, 4K) and various aspect ratios. Tip: use Google Search grounding to generate images of real-world landmarks or products, or use Minimal/High thinking mode for complex prompts.