Recently, it was confirmed that the highly anticipated Google flagship large model Gemini 3.5 Pro will officially launch on July 17th. This release date is extremely delicate—coinciding with the official release of DeepSeek V4, and the "competition" between these two top-tier large models is undoubtedly a major highlight in the AI industry.

At Google I/O in May, Gemini 3.5 Pro was originally scheduled to be released in June but was eventually postponed. According to insiders, this delay was not solely due to technical bottlenecks, but rather a strategic decision by the Google team: they chose to delay the release in order to skip fine-tuning the previous version Gemini 2.5 Pro and instead invest extra time in new pre-training. This decision aims to achieve a qualitative performance improvement through deeper computational power investment.

image.png

From the current leaked test information, the key improvements of Gemini 3.5 Pro focus on "front-end generation" capabilities. The model has made significant advances in UI design taste, concise code generation, and SVG vector graphics construction, with output results being more concise and accurate. In game development scenarios, the model also performs stably, capable of efficiently handling complex logic interactions.

Although this upgrade is unprecedented in scale, industry analysts believe that Gemini 3.5 Pro may still not be able to fully challenge models like Fable5 from Anthropic, which have parameters reaching trillions. However, Google has more cards up its sleeve. As a pioneer in the multimodal field, Google will simultaneously launch a new Nano Banana Pro image generation model based on the new Gemini 3.5 Pro foundation, with the core goal directly targeting the industry benchmark GPT-Image2, aiming to regain influence in the image generation niche market.

With Google's inherent advantage in the breadth of world knowledge, the practical performance of Gemini 3.5 Pro is highly anticipated. As July 17th approaches, the industry battle centered around computing power, training depth, and multimodal interaction is about to reveal its outcome.