Post Thumbnail

DeepSeek to open source its AI models

Chinese startup DeepSeek, which surprised Silicon Valley with the high performance of its AI models, has announced an unprecedented step – publishing key codes and data in open access. The company plans to start sharing its code repositories with all developers and researchers starting next week.

“We’re a small team exploring AGI. Starting next week, we’ll open access to 5 repositories, sharing our modest but sincere progress with full transparency,” the company stated in its X account.

The 20-month-old Hangzhou startup intends to go further than its competitors by providing access not only to models but also to base code, training data, and development methodology. This will allow anyone to download, modify, and improve the code underlying the highly rated R1 model and other company platforms.

DeepSeek’s decision strengthens the trend toward open AI development, which gained more supporters after the company’s models outperformed competitors from OpenAI and Meta in benchmark tests. Unlike OpenAI, which started as a partially open project but later departed from this policy, DeepSeek declares its intention to make all aspects of development transparent.

Company founder Liang Wenfeng, who previously ran a quantitative hedge fund, emphasized in a rare interview with Chinese media that the company does not prioritize commercialization of its AI models, seeing advantages in open source. “No ivory towers – just pure garage innovation energy and community-driven development,” the company stated.

This move could significantly impact the race between the US and China in developing advanced AI models. While investors have poured tens of billions of dollars into major American AI startups like Anthropic PBC and xAI, expecting significant returns, DeepSeek, which hasn’t disclosed external funding, can afford to focus less on building a revenue model.

DeepSeek has already forced larger competitors like Baidu to adopt the open-source concept. However, global players like OpenAI and Anthropic still keep their AI models, repositories, and data closed, making the Chinese startup’s move even more significant for industry development.

Experts note that DeepSeek’s open approach could accelerate AI technology development through collective efforts of developers worldwide, although this also raises security concerns from US and Australian governments.

Autor: AIvengo
For 5 years I have been working with machine learning and artificial intelligence. And this field never ceases to amaze, inspire and interest me.

Latest News

ChatGPT calls users "star seeds" from planet Lyra

It turns out ChatGPT can draw users into the world of scientifically unfounded and mystical theories.

AI music triggers stronger emotions than human music

Have you ever wondered why one melody gives you goosebumps while another leaves you indifferent? Scientists discovered something interesting. Music created by artificial intelligence triggers more intense emotional reactions in people than compositions written by humans.

GPT-5 was hacked in 24 hours

2 independent research companies NeuralTrust and SPLX discovered critical vulnerabilities in the security system of the new model just 24 hours after GPT-5's release. For comparison, Grok-4 was hacked in 2 days, making the GPT-5 case even more alarming.

Cloudflare blocked Perplexity for 6 million hidden requests per day

Cloudflare dealt a crushing blow to Perplexity AI, blocking the search startup's access to thousands of sites. The reason? Unprecedented scale hidden scanning of web resources despite explicit prohibitions from owners!

Threats and $1 trillion don't improve neural network performance

You've surely seen these "secret tricks" for controlling neural networks. Like threats, reward promises, emotional manipulations. But do they actually work? Researchers from the University of Pennsylvania and Wharton School conducted a large-scale experiment with 5 advanced models: Gemini 1.5 Flash, Gemini 2.0 Flash, GPT-4o, GPT-4o-mini and GPT o4-mini.