<ul><li>Chinese AI startup DeepSeek upgraded its R1-0528 reasoning model to compete with OpenAI and Google's top models.</li><li>The model saw significant improvements in reasoning, inference, accuracy, and increased token usage per question from 12,000 to 23,000.</li><li>DeepSeek's R1-0528 offers enhanced capabilities in math, programming, logic, function calling, and vibe coding.</li><li>The open-source model is free to use, can be customized, and is offered by Amazon Web Services and Microsoft Azure, with data staying in client-chosen servers.</li></ul>

DeepSeek Upgrades AI Reasoning Model to Rival OpenAI and Google

Discover more