Recently, a model dubbed 'gemini-3.8-flash' surfaced on the large model platform Arena. After thorough verification from multiple credible sources, it has been confirmed that this model is, in fact, Google’s next-generation flagship model, Gemini 4 Pro (with the internal codename Argon), currently undergoing testing phases. In several rounds of blind testing, Gemini 4 Pro showcased exceptional performance across tasks including SVG generation, 3D rendering, and intricate JSON processing, comprehensively surpassing competitors such as GPT-6 Astra and Claude Fable 5.1. For instance, when tasked with generating an SVG image of a 'pelican riding a bicycle,' it displayed meticulous handling of lighting, shadows, contours, and geometric relationships, highlighting its robust spatial reasoning abilities. Furthermore, Gemini 4 Pro particularly excelled in coding, agent-based tasks, and extended-duration tasks, with its performance advantages becoming increasingly evident as both task duration and complexity rose. Currently, Google has not officially launched Gemini 4 Pro, but it is anticipated to make its formal debut in October 2026, further bolstering Google’s competitive edge in the large model market.