A Low-Cost, Open-Source AI Model Comparable to ChatGPT

By Dr. Shripad Bhat
January 29, 2025

Training Large Language Models (LLMs) requires huge computing power and resources. Reportedly, the cost of training GPT-4 was more than $100 million (Link: https://www.wired.com/story/openai-ceo-sam-altman-the-age-of-giant-ai-models-is-already-over/ ).

Currently, there are sanctions against exporting advanced chips to China (Link:https://en.wikipedia.org/wiki/United_States_New_Export_Controls_on

_Advanced_Computing_and_Semiconductors_to_China). Without access to such resources, a Chinese startup built an efficient, low-cost model with a training cost of only $ 6 million!. The performance of this model is roughly comparable to OpenAI o1. Impressive, right?

While cutting edge models like GPT-4 (owned by OpenAI) are not open source, DeepSeek has been released as open source.

Code for embedding this tweet in blogpost)

(<blockquote class=”twitter-tweet”>🚀 DeepSeek-R1 is here! ⚡ Performance on par with OpenAI-o1 📖 Fully open-source model & technical report 🏆 MIT licensed: Distill & commercialize freely! 🌐 Website & API are live now! Try DeepThink at <a href=”https://t.co/v1TFy7LHNy”>https://t.co/v1TFy7LHNy</a> today! 🐋 1/n <a href=”https://t.co/7BlpWAPu6y”>pic.twitter.com/7BlpWAPu6y</a>— DeepSeek (@deepseek_ai) <a href=”https://twitter.com/deepseek_ai/status/1881318130334814301?ref_src=twsrc%5Etfw”>January 20, 2025</a></blockquote> <script async src=”https://platform.twitter.com/widgets.js” charset=”utf-8″></script>

Link to the full technical report of DeepSeek: https://github.com/deepseek-ai/DeepSeek-R1/blob/main/DeepSeek_R1.pdf