Alibaba claims its new AI model can beat DeepSeek, ChatGPT
What's the story
Alibaba has announced the launch of its latest artificial intelligence (AI) model, Qwen 2.5 Max. The Chinese tech giant's cloud division claims that this new model outperforms leading AI models like DeepSeek, OpenAI's GPT-4o, and Meta's Llama. Alibaba's introduction of Qwen 2.5 Max comes as a strategic response to the rise of DeepSeek. The latter recently launched the powerful R1 model, which has created quite a stir in the tech world for being cost-effective and energy efficient than its rivals.
Features
Take a look at the benchmark performance
Qwen2.5-Max was assessed alongside top models on various benchmarks, including MMLU-Pro (tests knowledge with college-level problems), LiveCodeBench (evaluates coding skills), LiveBench (tests general capabilities), and Arena-Hard (approximates human preferences). As per the company, Qwen2.5-Max beat DeekSeek's V3 model and Meta's Llama 3.1-405B across all metrics. Qwen2.5-Max is available in Qwen Chat, where you can interact directly with the model, explore artifacts, search, and more.
Pricing competition
AI price war intensifies as companies vie for market share
The release of DeepSeek's V2 model last year had sparked a price war in China's AI market. The open-source model, priced at just 1 yuan (₹11) per 1 million tokens processed, was way cheaper than most of its competitors. This aggressive pricing prompted Alibaba's cloud division to slash its prices by up to 97% on some models, further intensifying the competition among giants including Tencent and Baidu.
Industry recognition
OpenAI acknowledges DeepSeek's success but raises concerns
OpenAI CEO Sam Altman recently acknowledged DeepSeek's success, calling its progress "impressive." However, the company has now expressed concerns that competitors, inlcuding DeepSeek, are leveraging its work to quickly advance their own AI tools. Meanwhile, Microsoft is investigating a potential case of improper data acquisition involving DeepSeek, according to Bloomberg.