Deepseek News: China’s high-tech company, Alibaba, has released a new version of the QWEN 2.5 artificial intelligence model that claimed that it has exceeded the very high reputation on Wednesday, said Royter, reportedly reported.
Most Chinese are off work and the unusual timing of the release of QWEN 2.5-Max is pointing out that the AI Startup DeepSeek’s meteor in the past three weeks has increased. Not only overseas rivals, but also domestic competition.
“QWEN 2.5-Max is excellent … GPT-4O, Deepseek-V3, LLAMA-3.1-405B” and Alibaba’s cloud units refer to Openai and Meta’s most advanced open openings. , In the announcement posted on the official WeChat account. -AI model source.
In addition to Tencent Holdings LTD. and Baidu Inc., Alibaba has poured considerable resources into a cloud service segment, and is engaged in hot contests to recruit Chinese AI developers and use tools.
Deepseek, a 20 -month startup established in Hangzhou in Alibaba’s Home City, sent a shock wave this week. Alibaba Cloud has also shared a score that suggests that AI will defeat Openai and Anthropic models on a specific benchmark.
Cloud service providers such as Alibaba and Tencent have significantly reduced prices in recent months to acquire more users. DeepSeek has already contributed to the price war, along with other promising AI startups in China, which secured funds for unicorn evaluation.
Deepseek R1 rattles high -tech companies
DeepSeek has shaken things in the AI world. The AI Assistant with the Deepseek-V3 model was released on January 10, followed by the R1 model on January 20. It is really worth it considering that DeepSeek is said to have low development costs and use costs.
Returning to his hometown, the success of the startup has begun a fierce competition for high -tech peases in China. Only two days after the R1 model was dropped, the ordinance of Tiktok’s parent company counterattacks with its own flagship AI model, claiming that it is better than O1’s O1’s Microsoft AI betting. Instructions in an important benchmark test that understands the complex complexity of AI and measures how to react.
Of course, DeepSeek has already made the same, stating that R1 models can move directly using O1 of OPENAI on multiple performance benchmarks.
(Reuters are input from Bloomberg)
