Nano Banana Pro, Google's Gemini 3 Pro Image Preview, posts an intelligence index of 48.4 and a coding index of 46.5, placing it among the stronger models in the current field. GPQA at 0.908, livecodebench at 0.917, and scicode at 0.561 are all high, and the agentic index of 64.4 supports reliable multi-step tool use. Vision and image output capability add meaningful multimodal breadth.

For businesses, this model suits demanding coding tasks, visual content workflows, and agentic automation. The livecodebench score in particular is among the highest in this batch. The primary limitation is the 65K context window, which is restrictive for long-document analysis or large codebase tasks compared to competitors offering 400K or 1M tokens.

At $2.00 input and $12.00 output, pricing is on the higher end. Teams prioritising coding quality and multimodal capability over context depth will find this a strong performer, but those needing large context should look to Google's Gemini 3 Pro Preview with its 1M token window.

Assessed June 6, 2026

Rankings consider pricing, capabilities, benchmarks, and real-world applicability and are refreshed as new models launch. Feedback?