Back to today’s briefing

Story perspectives

DeepSeek's AI Model Challenges Giants with Cost-Effective Edge

1/2/2026

31 7 Full Breakdown

Summary

Story summary

DeepSeek, a Chinese AI startup, released a free app and the DeepSeek V3 model, reportedly outperforming Meta's Llama 3.1 and rivaling OpenAI's GPT-4 while using less computing power.
DeepSeek was founded in April 2023 by Liang Wenfeng, founder of DeepSeek, and backed by a hedge fund.
Training DeepSeek V3 reportedly cost about $560 million.
The cost advantage could disrupt the AI market, prompting firms to lower fees to retain users.

1 of 1

Story summary

DeepSeek, a Chinese AI startup, released a free app and the DeepSeek V3 model, reportedly outperforming Meta's Llama 3.1 and rivaling OpenAI's GPT-4 while using less computing power.
DeepSeek was founded in April 2023 by Liang Wenfeng, founder of DeepSeek, and backed by a hedge fund.
Training DeepSeek V3 reportedly cost about $560 million.
The cost advantage could disrupt the AI market, prompting firms to lower fees to retain users.