**通义千问发布Qwen1.5系列首个千亿参数开源模型：Qwen1.5-110B**
近日，通义千问团队宣布推出Qwen1.5系列的首个千亿参数开源模型——Qwen1.5-110B。这一模型在人工智能领域引起广泛关注，标志着自然语言处理技术的又一重要突破。
据悉，Qwen1.5-110B模型拥有高达千亿参数，采用Transformer解码器架构，与其他Qwen1.5模型一脉相承。该模型在基础能力评估中表现出色，与Meta-Llama3-70B相媲美。尤其在Chat评估中，其表现尤为突出，通过了MT-Bench和AlpacaEval 2.0的严格测试。
值得一提的是，Qwen1.5-110B模型支持多语言处理，包括英语、中文、法语、西班牙语、德语、俄语、日语、韩语、越南语以及阿拉伯语等。此外，该模型还支持高达32K tokens的上下文长度，极大提升了模型推理时的效率。这一创新得益于其采用的分组查询注意力（GQA）机制。
业内专家表示，Qwen1.5-110B的发布对于自然语言处理领域具有重大意义，它不仅提升了模型的性能，还为未来的语言模型发展提供了新的思路。通义千问团队通过开源这一模型，为科研人员和开发者提供了一个强大的工具，期待其在各个领域的应用能够带来实质性的进步。
此消息的来源为Qwen官方通报。随着模型的开源，预计将引发新一轮的技术研究与探讨。
英语如下：
News Title: “TongyiQianWen releases Qwen1.5-110B model: A beast with over 10 billion parameters, leading the new era of multi-language AI chatting globally!”
Keywords: TongyiQianWen, Qwen1.5 series, open-source model
News Content: **TongyiQianWen team releases the first open-source model with over 10 billion parameters in the Qwen1.5 series: Qwen1.5-110B**
Recently, the TongyiQianWen team announced the launch of Qwen1.5-110B, the first open-source model with over 10 billion parameters in the Qwen1.5 series. This model has attracted widespread attention in the AI field, marking another important breakthrough in natural language processing technology.
It is reported that the Qwen1.5-110B model has up to 10 billion parameters, using the Transformer decoder architecture and is closely related to other Qwen1.5 models. The model has demonstrated excellent performance in basic ability assessments, comparable to Meta-Llama3-70B. Especially in Chat assessments, its performance is particularly prominent, passing strict tests by MT-Bench and AlpacaEval 2.0.
It is worth mentioning that the Qwen1.5-110B model supports multi-language processing, including English, Chinese, French, Spanish, German, Russian, Japanese, Korean, Vietnamese, and Arabic. In addition, the model supports a context length of up to 32K tokens, greatly improving the efficiency of model inference. This innovation is due to its adopted Grouped Query Attention (GQA) mechanism.
Industry experts indicate that the release of Qwen1.5-110B is of great significance to the field of natural language processing. Not only does it enhance the performance of the model, but it also provides new insights for the development of future language models. Through open-sourcing this model, the TongyiQianWen team provides a powerful tool for researchers and developers, and it is expected to bring substantial progress in various fields.
The source of this news is from Qwen’s official announcement. With the open-sourcing of the model, it is expected to trigger a new round of technical research and discussion.
【来源】https://qwenlm.github.io/zh/blog/qwen1.5-110b/
Views: 1