据The Information报道，Meta的下一代大型语言模型Llama 4目前仍处于开发阶段，但在某内部基准测试中的表现并不理想，特别是在数学处理能力和推理能力方面，Llama 4仍需进行大量改进。

有外媒指出，由于竞品DeepSeek等带来的竞争压力，Meta正在加紧对Llama 4进行改进，以提升其性能。据悉，Meta计划将借鉴DeepSeek的机器学习方法，为Llama 4引入逻辑思考体系，同时还会改善模型本身的语音输出效果。

尽管Llama 4还需要更多时间进行优化，但外媒指出，由于投资行业的压力，Meta仍打算在本月晚些时候推出Llama 4。

此前，Meta曾表示计划在4月29日举办LlamaCon活动，旨在帮助开发者利用Llama构建应用和产品。预计Meta会在此次活动中正式发布Llama 4。

On April 5th, The Information reported that Meta’s next-generation large language model, Llama 4, is still under development. However, its performance in an internal benchmark test was not ideal, especially in terms of mathematical processing and reasoning abilities. Llama 4 requires significant improvements.

Due to the competitive pressure from rival products like DeepSeek, Meta is stepping up efforts to enhance Llama 4’s performance. It is reported that Meta plans to incorporate machine learning methods inspired by DeepSeek into Llama 4, introducing a logical thinking system and also improving the model’s own voice output effects.

Despite requiring more time for optimization, Meta still intends to launch Llama 4 later this month, as per reports citing pressure from the investment community.

Previously, Meta had announced plans to host the LlamaCon event on April 29th, aiming to help developers build applications and products using Llama. The official release of Llama 4 is expected to take place during this event.