AI meets jugaad as Alibaba's Qwen 2.5 pips DeepSeek in model battle
DeepSeek V3, Chinese tech giant Alibaba on Tuesday released a new version of its Qwen 2.5 AI model.
Budget with ET
How Budget 2025 can walk the tightrope of India's 'sins'
Will Budget add more firepower to defence industry's local manufacturing?
Sitharaman and co to fuel India's luxe aspirations, premium growth?
Qwen achieves competitive performance against top-tier models and outcompetes DeepSeek V3 in popular coding and user query benchmarks, Alibaba said, adding that it will continue scaling the model in pre-training and further invest in boosting its reinforcement learning to allow for improved reasoning.
Deepseek, Alibaba, Tencent and other Chinese companies have figured out a way to make do with the fewer resources at their disposal amid US sanctions on exporting AI chips to China. The result? Open-source AI models that match ChatGPT creator OpenAI’s model in reasoning benchmarks, while delivering results nearly twice as fast, trained at less than a tenth of compute resources.
Also Read: ETtech Explainer: The story behind DeepSeek’s greener and leaner chatbot
ET explains how the Chinese have marvelled cost-cutting for AI training:
Artificial Intelligence(AI)
Java Programming with ChatGPT: Learn using Generative AI
By — Metla Sudha Sekhar, IT Specialist and Developer
Artificial Intelligence(AI)
Basics of Generative AI: Unveiling Tomorrows Innovations
By — Metla Sudha Sekhar, IT Specialist and Developer
Artificial Intelligence(AI)
Generative AI for Dynamic Java Web Applications with ChatGPT
By — Metla Sudha
