On January 20, a China-based startup, DeepSeek, made waves in artificial intelligence (AI) circles with the launch of its R1 model. The day after its release, chipmaker Nvidia saw its market value drop by approximately $600 billion in a single day albeit with a big rebound the next day.
DeepSeek’s AI Assistant has surpassed OpenAI’s ChatGPT to become the most downloaded free app on Apple’s App Store and Google Play Store in the US. Based on public data from Google Play Store and App Store, Botreview estimates that at least 3.4 million users have downloaded the app in the last 7 days (at time of posting).
An analysis from SEO.ai indicates that DeepSeek has between 5 to 6 million active users. The main markets are China (23%), the United States (15%) and Egypt (6%).
- DeepSeek claims that its latest R1 model was developed for less than $6 million. If true, that means that training DeepSeek’s R1 was 20 times cheaper than what’s typically required for flagship models like OpenAI’s GPT series.
- Performance of R1 is on par with OpenAI-o1, Claude 3.6 and Gemini-1206 (according to DeepSeek’s whitepaper and third party testers)
- The price per token is 1/30th of similar OpenAI models, costing $2.19 per million output tokens versus OpenAI’s 01 model at $60.00.
- Due to chip export restrictions to China, DeepSeek achieved its claimed results using less-advanced chips, challenging the belief that cutting-edge chips from companies like Nvidia are necessary for high-performing AI models.
- DeepSeek-R1 is free to use and open source.