Comparative Analysis of AI Assistants: Testing DeepSeek, ChatGPT

DeepSeek's R1 model, powered by Nvidia's H800 chips, offers an affordable alternative to more expensive AI models, with its development costing $5.6 million compared to OpenAI's significantly higher expenditures. Despite its lower cost, DeepSeek R1 claims comparable performance to OpenAI's models, even outperforming some versions in specific tests, although its MMLU benchmarks may not fully reflect real-world capabilities. The model faces challenges in handling politically sensitive topics in China, raising concerns about censorship and the influence on AI development, a broader issue highlighted by the Stargate LLM project. Researchers have widely adopted DeepSeek-R1 for scientific research in fields like mathematics and cognitive neuroscience, benefiting from its affordability and accessibility. While its scientific task performance is on par with OpenAI's o1 model, it remains a cost-effective option with over three million downloads, demonstrating its popularity despite solving only a fraction of real scientific problems accurately.