Alibaba Unveils New AI Model, Claims It Outperforms DeepSeek & GPT-4o

Alibaba Unveils New AI Model, Claims It Outperforms DeepSeek & GPT-4o

SUMMARY

Alibaba unveiled Qwen 2.5-Max and claimed it outperformed DeepSeek-V3, GPT-4o and Llama-3.1-405B

The launch came just days after Chinese AI startup DeepSeek rolled out a free AI assistant, which it claims uses less data

While DeepSeek-V3 model was trained using NVIDIA H800 chips at a cost of $6 Mn, OpenAI’s GPT-4 model was trained using 100K NVIDIA GPUs at a cost of around $100 Mn

Chinese tech giant Alibaba has rolled out a new version of its Qwen 2.5 artificial intelligence (AI) model, claiming that it outperformed DeepSeek’s most advanced open source AI model that has taken the world by storm.

The company’s cloud computing unit Alibaba Cloud said in its official WeChat account that Qwen 2.5-Max surpassed OpenAI’s GPT-4o, DeepSeek-V3 and Meta’s Llama-3.1-405B ‘almost’ across the board, Reuters reported.

The move comes at a time when Chinese AI startup DeepSeek has shattered beliefs about the cost of AI, putting immense pressure on leading AI firms like OpenAI and Meta.

Last week, DeepSeek launched its free AI assistant DeepSeek-V3, claiming that it uses less data and costs only a fraction of services offered by incumbents.

DeepSeek researchers wrote in a paper last month that the DeepSeek-V3 model, released on January 10, used NVIDIA’s H800 GPUs for training, and cost less than $6 Mn. In comparison, OpenAI’s GPT-4 model was trained using 100K NVIDIA GPUs (the more advanced H100) at the cost of $100 Mn.

The fact that DeepSeek APIs cost up to 95% less than OpenAI’s latest models prompted investors to question the huge spending by AI behemoths in the US.

The emergence of the low-cost Chinese AI model and its rising popularity on app stores worried investors, setting a selloff in tech stocks such as NVIDIA, Broadcom, Alphabet and Microsoft earlier this week.

DeepSeek has also released an AI model called Janus-Pro-7B for AI-led image generation, which it claims outperforms OpenAI’s DALL-E 3 and Stability AI’s Stable Difussion. This is likely to put further strain on tech stocks.

However, for Indian startups looking to build their own large language models from scratch, DeepSeek’s meteoric rise is good news. While founders of AI startups in India have thus far lamented about the need of intensive capital to build LLMs, DeepSeek has shattered these beliefs.

Recently, AI search engine Perplexity’s founder and CEO Aravind Srinivas said that India can build cheaper AI models through ingenuity and strategic investments in open source models.

“I think that’s possible for AI (to train models frugally), given the recent achievements of DeepSeek. So, I hope India changes its stance from wanting to reuse models from open-source and instead trying to build muscle to train their models that are not just good for Indic languages but are globally competitive on all benchmarks,” Srinivas said.

 

Note: We at Inc42 take our ethics very seriously. More information about it can be found here.

You have reached your limit of free stories
Become A Startup Insider With Inc42 Plus

Join our exclusive community of 10,000+ founders, investors & operators and stay ahead in india's startup & business economy.

2 YEAR PLAN
₹19999
₹7999
₹333/Month
UNLOCK 60% OFF
Cancel Anytime
1 YEAR PLAN
₹9999
₹4999
₹416/Month
UNLOCK 50% OFF
Cancel Anytime
Already A Member?
Discover Startups & Business Models

Unleash your potential by exploring unlimited articles, trackers, and playbooks. Identify the hottest startup deals, supercharge your innovation projects, and stay updated with expert curation.

Alibaba Unveils New AI Model, Claims It Outperforms DeepSeek & GPT-4o-Inc42 Media
How-To’s on Starting & Scaling Up

Empower yourself with comprehensive playbooks, expert analysis, and invaluable insights. Learn to validate ideas, acquire customers, secure funding, and navigate the journey to startup success.

Alibaba Unveils New AI Model, Claims It Outperforms DeepSeek & GPT-4o-Inc42 Media
Identify Trends & New Markets

Access 75+ in-depth reports on frontier industries. Gain exclusive market intelligence, understand market landscapes, and decode emerging trends to make informed decisions.

Alibaba Unveils New AI Model, Claims It Outperforms DeepSeek & GPT-4o-Inc42 Media
Track & Decode the Investment Landscape

Stay ahead with startup and funding trackers. Analyse investment strategies, profile successful investors, and keep track of upcoming funds, accelerators, and more.

Alibaba Unveils New AI Model, Claims It Outperforms DeepSeek & GPT-4o-Inc42 Media
Alibaba Unveils New AI Model, Claims It Outperforms DeepSeek & GPT-4o-Inc42 Media
You’re in Good company