DeepSeek R1 just made reasoning a commodity
R1 roughly matches OpenAI o1 on math benchmarks, costs about 27x less per output token and ships as open weights. Plan your AI roadmap for falling model prices.
R1 roughly matches OpenAI o1 on math benchmarks, costs about 27x less per output token and ships as open weights. Plan your AI roadmap for falling model prices.
OpenAI's Operator, the $500B Stargate project and the Musk-Altman feud filled the headlines, but DeepSeek R1's reasoning models were the story that mattered.
Small language models are faster, cheaper and run on laptops and phones. Ten real uses, from blocking text scams onward, where a smaller model beats a frontier one.
Nvidia's entry into personal AI computing, new AI privacy concerns and Bill Gates on the future of work: the second weekly AI digest of January 2025.
Databricks' 2024 report: 11x more ML models in production, 377% growth in vector databases, 70% of companies using RAG. The numbers an executive should know.
The first weekly AI digest of 2025: DeepSeek's rise as China's AI contender, platform shifts, Google search quality concerns and other industry moves.
The AI trends shaping 2025: companies treating AI as augmentation rather than replacement, the infrastructure race, and what leadership teams should plan for.