Search AI Products and News

On the first day of DeepSeek Open Source Week, we officially released our latest technological achievement, FlashMLA, an efficient Multi-Layer Attention decoding kernel specifically designed for NVIDIA Hopper architecture GPUs. This technology is optimized for variable-length sequence scenarios and can significantly enhance the inference performance of large models. The core technical features of FlashMLA include comprehensive support for BF16 precision and a paged key-value cache with a block size of 64.

Recently, Huggingface, the world's largest open source AI community, released the latest Open LLM Leaderboard. The results show that all ten of the top-ranked open source large models are derivative models that have been fine-tuned based on Alibaba Tongyi Qwen's (Qwen) open-source model. This achievement marks Qwen's dominant position in the open-source AI field and further enhances its global influence. The Open LLM Leaderboard is widely...

Recently, the DeepSeek team in China has launched its latest open-source large model R1, which has gained widespread attention. The performance of the R1 model is exceptionally outstanding, surpassing OpenAI's o1 model in multiple tests, especially excelling in assessments related to mathematics and programming. In the latest AIME2024 test in the United States, R1 scored 79.8, outperforming o1's score of 79.2. In the MATH-500 test, R1 achieved a score of 97.3, again leading over o1.

The 2024 Yunqi Conference opened grandly in Hangzhou, with Alibaba Group CEO and Alibaba Cloud Intelligence Group Chairman and CEO Wu Yongming delivering a keynote speech. During the conference, Alibaba announced a series of groundbreaking releases, including the launch of the next-generation open-source model Qwen2.5, which outperforms Llama405B and firmly holds the position of the 'strongest open-source large model'.

Shenzhen Yuanxiang Information Technology Co., Ltd. recently announced the successful release of China’s largest Mixture of Experts (MoE) open source large model — XVERSE-MoE-A36B. The launch of this model marks a significant advancement in China’s AI field, elevating domestic open source technology to an internationally leading level.

At the 51st SIGGRAPH conference, NVIDIA CEO Jensen Huang and Meta CEO Mark Zuckerberg engaged in a fireside chat discussing the future of artificial intelligence and simulation. During their conversation, Huang praised Meta's open source large model strategy and emphasized the importance of open source for AI development. They showcased their friendship and shared vision for advancing the industry through a creative 'coat swap' ceremony.

China's technology giant Alibaba's latest open-sourced Qwen-2-72B fine-tuned version has topped the global open source large model ranking.

Google has released the open source AI large model Gemma, with the domestic open source large model InternLM2 outperforming Gemma in all aspects. The competition in the open source large model field is intense, making the large model landscape increasingly lively.

Yann LeCun defended the Llama open source large model in the Senate, emphasizing the importance of open source. Artificial intelligence should be viewed as infrastructure, and open source can facilitate its rapid development. Open source models help democratize access to cutting-edge technology and reduce social disparities. There is an urgent need for collaboration to establish standards and governance models for artificial intelligence. Yann LeCun supports related legislation and is willing to cooperate with relevant personnel within companies.

The first batch of large models' industry implementation progress in the country has come to light, and open source large models are bringing changes to the market landscape. The practical implementation of large models is initially showing potential and transformation in industries such as energy, finance, and pharmaceuticals. Enterprises are flocking to open source large models, while major companies are adjusting their strategies, with one hand on closed source and the other on open source.

With mathematical capabilities surpassing ChatGPT, Claude Instant 1, and PaLM 2-540B, the WizardMath model developed by Microsoft's all-Chinese team has gained popularity for its ability to enhance large model capabilities generated by AI instructions.