Jev is new and can make decisions based on pictures!

AI Application Wind Vane (Public Account: ZhidxcomAI) Compiler｜Bi Weihao Editor｜Li Shuiqing Zhidongxi reported on October 15th that this morning, the American AI company TypeSafe AI released the new model Jev 1.14.0. At the same time, it released three experimental functions: image input, new problem type Box (non-decision-making problem), and 262k long context. The three functions were unified into a group called "Danger Zone" (Danger Zone). Zone) experimental plan. These three functions are turned off by default, requiring users to manually turn on the API switch. The official statement stated that these functions have not reached production standards and are at their own risk. On September 16, TypeSafe, which has been in stealth development for two years, officially launched Jev. This model is different from large language models that generate continuous text. It directly outputs structured decisions such as classification, options, right and wrong judgments, and provides calibrated confidence. TypeSafe calls it the first System One model, which refers to a decision-making model that is closer to fast, intuitive judgment. The release of Jev allows decision-making models to quickly enter the field of vision of more people. On October 10, TypeSafe announced the completion of a US$870 million (approximately RMB 5.86 billion) Series A financing, with a valuation of US$7.5 billion (approximately RMB 50.5 billion), led by a16z, with participation from Sequoia Capital and DCVC. As this field received attention, Microsoft launched Microsoft-Decision-1 to support structured decision-making, and Jev projects such as Kev, Laya, and Decider emerged in the open source community. The release of Jev 1.14.0 is also the first model update after TypeSafe announced its financing. 1. The model has been slightly upgraded, and the official reminds us not to trust the evaluation Jev 1.14.0 itself is not a major upgrade. TypeSafe describes it as a "non-technical" update, with subtle changes but wide coverage. The most obvious improvement is expected to be multi-language capabilities. However, Chinese users cannot expect it to maintain the same performance as English for the time being.