News
Google Strikes hard: Gemini 3
2 min read
Source: Telegram AI频道
Google strikes hard: Gemini 3.5 Pro is locked and released on July 17th, a full confrontation with DeepSeek V4. Recently, it was confirmed that the much-anticipated Google flagship large model Gemini3.5Pro will be officially unveiled on July 17th. This release node is extremely subtle - just as the official version of DeepSeek V4 is released, the "competition" between two top large models will undoubtedly become a major attraction in the AI industry. At the previous Google I/O conference in May, Gemini3.5Pro was originally scheduled to be released in June, but was eventually postponed for some reason. According to insiders, the delay was not simply a technical bottleneck, but a strategic decision made by the Google team: they would rather postpone the release than skip fine-tuning the old version of Gemini2.5Pro and invest extra time in new pre-training. This decision aims to bring qualitative changes in model performance through deeper investment in computing power. Judging from the test information currently flowing out, the improvement of Gemini3.5Pro focuses on the "front-end generation" capability. This model has made significant leaps forward in terms of UI design taste, streamlined code generation, and SVG vector graphics construction, and the output results are more capable and accurate. In game development scenarios, the model performance is equally robust and can handle complex logical interactions more efficiently. Although this upgrade is unprecedented, industry analysts believe that Gemini3.5Pro may not be able to completely shake "Big Mac" models with trillion-level parameters such as Anthropic's Fable5 in terms of parameter scale. However, Google has more up its sleeve than that. As a pioneer in the multi-modal field, Google will simultaneously launch the new Nano Banana Pro image generation model based on the new Gemini3.5Pro base. Its core goal is directly aimed at the current industry benchmark GPT-Image2, with the intention of regaining the right to speak in the segmented track of image generation. With Google’s inherent advantage in the breadth of the world’s knowledge base, the actual performance of Gemini3.5Pro is highly anticipated. As July 17 approaches, the answer to this industry war surrounding computing power, training depth, and multi-modal interaction is about to be revealed. via AI News (author: AI Base)