据科技媒体The Information报道，Meta计划于今年7月发布其最新的大模型Llama 3。该模型旨在与OpenAI的GPT-4相竞争，后者因其强大的多模态功能而备受瞩目。尽管Meta希望Llama 3能够达到与GPT-4相媲美的能力，但据内部人士透露，由于尚未开始对模型进行微调，公司尚未决定Llama 3是否将成为多模态模型。此外，与去年发布的Llama 2相比，Llama 3的参数预计将显著增加，最高可达1400亿个参数，而Llama 2的参数最高为700亿。在推出Llama 3之前，Meta还致力于解决Llama 2存在的问题，特别是在处理争议性问题方面的不足。
Title: Meta to Launch Llama 3 as Competitor to GPT-4
Keywords: Meta, Llama 3, GPT-4
News content: Tech media outlet The Information reports that Meta plans to release its latest large model, Llama 3, in July. The model is intended to compete with OpenAI’s GPT-4, which has gained attention for its robust multimodal capabilities. While Meta hopes Llama 3 will match the capabilities of GPT-4, an insider has revealed that the company has yet to decide if Llama 3 will be multimodal as it has not started fine-tuning the model. Additionally, Meta is expected to significantly increase the number of parameters in Llama 3, with a projected maximum of 1400 billion, compared to Llama 2’s maximum of 700 billion. Before launching Llama 3, Meta is also working to address issues identified in Llama 2, particularly its inability to handle controversial topics.
【来源】https://wallstreetcn.com/articles/3709312
Views: 1