Gemini 3 Flash (simonwillison.net)

Google has introduced the Gemini 3 Flash model, which significantly enhances performance while being more cost-effective than its predecessor, Gemini 3 Pro. Priced at a quarter of the cost of the 3 Pro, Gemini 3 Flash performs exceptionally well across benchmarks and supports a wide range of input formats, including text, image, and video. It maintains identical technical specifications to the 3 Pro, allowing for maximally efficient processing with an input capacity of over one million tokens and a comparable knowledge cut-off date of January 2025. This release is particularly noteworthy for the AI/ML community because it not only offers improved performance but also introduces a new tier of thinking levels for generating outputs, enhancing its versatility for developers. However, the absence of image segmentation capabilities marks a setback for users relying on this feature, encouraging them to continue using the Gemini 2.5 series for more specialized tasks. Overall, Gemini 3 Flash represents an exciting advancement in accessible AI technology, allowing users to leverage its coding skills and multimodal functionality at an attractive price point.