Google's gemini 4 pro model shows strong svg image generation capabilities in testing

According to netizens, Google is currently testing its most advanced model, Gemini 4 Pro, under the codename 'gemini-3.8-flash,' on platforms like Arena. This approach, using an older version number, is a common method for quietly refining the model's capabilities. In shared test images, the model's ability to generate an SVG image of a 'pelican riding a bicycle' stood out. The output was deemed very excellent, with comparisons made to OpenAI's GPT-6Astra Pro on the same task. SVG, a vector format, is particularly challenging for models due to its requirement for understanding structure, paths, and geometric relationships. Netizens believe that once officially released, Gemini 4 Pro will demonstrate even greater capabilities, as the current testing version is not the final product. Google has not yet officially announced the model. The testing phase highlights the model's potential in graphic generation, especially with complex SVG formats. The comparison to GPT-6Astra Pro underscores the competitive landscape in AI image creation. While the model is still in testing, its performance suggests significant advancements in the field. Netizens are optimistic about its future release, anticipating enhanced capabilities beyond the current testing version. Google's decision to keep the model under wraps reflects a strategic approach to refining its features before public release.