AsianFin -- China Telecom's Artificial Intelligence Research Institute (TeleAI) has officially launched its "Complex Reasoning Large Model," TeleAI-t1-preview.
The model utilizes reinforcement learning methods and incorporates thinking paradigms such as exploration and reflection, significantly improving accuracy in complex problems like mathematical derivations and logical reasoning.
In the prestigious mathematical benchmarks of the U.S. Math Competition AIME 2024 and MATH500, TeleAI-t1-preview scored 60 and 93.8, respectively, surpassing benchmark models like OpenAI's o1-preview and GPT-4. In the graduate-level question-answering test, GPQA Diamond, TeleAI-t1-preview outperformed GPT-4 and rivaled the performance level of Claude 3.5 Sonnet.
Explore more exclusive insights at nextfin.ai.
