recently,360Zhinao officially announced that its independently developed "Qiyuan large model" has been filed and successfully implemented. It is worth mentioning that in September this year, the model was approved to be open to the public, making 360 the first domesticLarge ModelAll of them are technology companies that have passed the registration.
It is understood that the 360 Intelligent Brain Big Model has ten core capabilities, including generative creation, multi-round dialogue, and logical reasoning, covering all application scenarios of the big model. The model integrates the technical capabilities of the 360GPT big model and the 360 multimodal big model, and has been fully connected to 360 search, 360 browser and other Internet end applications.
It is worth mentioning that in August this year, Shanghai Jiao Tong University, Tsinghua University and the University of Edinburgh jointly launched C-Eval, a multi-level and multi-disciplinary Chinese evaluation suite suitable for large language models.up to dateIn the Chinese big model capability rankings, the 100-billion-parameter version of the 360 Intelligent Brain big model, 360GPT-S2, has an average score that exceeds GPT-4, demonstrating its strong strength.
In the evaluation results, the 360GPT-S2, a 100 billion parameter version of the 360 Intelligent Brain model, scored 82 points in social science questions, nearly 5 points higher than GPT-4; and scored 70.6 points in humanities questions, 6 points higher than GPT-4. These data fully demonstrate the superiority of the 360 Intelligent Brain model in handling complex problems and multi-field applications.