With the assistance of a newly released DeepSeek R1, which costs about 95% less than OpenAI’s competing models, some brilliant performances in intricate reasoning tasks had begun to be seen. The model was built for only \$5.6 million, which is far less than any amount that major IT companies spend on typical AI model development.
Early users have reported that DeepSeek R1 has lived up to its claim. Software developers have praised the model for generating highly optimized code, while others have confirmed success on tasks ranging from web development to chess strategy. Evaluation tests show that this model applies very well to mathematics and code challenges.
Also Read: FO rejects ‘baseless allegations’ targeting Pakistan’s commitment to One-China policy
DeepSeek can also perform on low-cost hardware, handling rather advanced tasks. Users then went on social media to give examples of what DeepSeek can do, showing its efficiency on regular consumer hardware for carrying out sophisticated computational tasks.
“In a recent interview, Liang Wenfeng, the founder of DeepSeek and a former hedge fund manager, said, “This is about democratizing innovation.” The company appears to be winning users as it charges developers less than $0.14 per million tokens compared to OpenAI’s $7.5 while offering free access to individual users.
DeepSeek’s achievements are particularly significant because its access to cutting-edge American semiconductor technology was severely restricted. When export controls were passed, the company managed to hoard 10,000 Nvidia H100 chips, allowing it to continue development despite the political strife.