Deep Reasoning Large Models

April 21 news, KU Xunfei today announced that Xunfei Starfire X1 new upgrade, said to be "the industry's only deep inference based on the national arithmetic training model". This upgrade has the following key information: The effect of math, code, logical reasoning, text generation, language understanding, knowledge quiz and other general tasks has been significantly improved, and in the case of the model parameters are one order of magnitude smaller than similar models in the industry, the overall effect is comparable to that of OpenAI o1 and DeepSeek R1. Incorporating more complex types of data into the scenario, the model's generalizability has made progress, and the model's generalization has been improved. ...

March 3 news, KU Xunfei today announced the upgrade of Starfire X1, and released a series of new products such as Starfire All-in-One. Test collection sources: Chinese test set from the 2023/2024 examination questions / simulation questions / competition questions, English test set from AIME 2024/2025 and MATH 500 competition set Among them, the depth of reasoning large model Starfire X1 upgraded again, mathematical ability to comprehensively against the DeepSeek R1 and OpenAI o1. KU Xunfe said the upgrade in the numb ... ...