DeepSeek’s updated R1 AI model is more censored, test finds

Chinese AI startup DeepSeek’s newest AI model, an updated version of the company’s R1 reasoning model, achieves impressive scores on benchmarks for coding, math, and general knowledge, nearly surpassing OpenAI’s flagship o3. But the upgraded R1, also known as “R1-0528,” might also be less willing to answer contentious questions, in particular questions about topics the […]

  • Related Posts

    Yoshua Bengio launches LawZero, a nonprofit AI safety lab

    Turing Award winner Yoshua Bengio is launching a nonprofit AI safety lab called LawZero to build safer AI systems, he told the Financial Times on Monday. LawZero raised $30 million…

    Continue reading
    Inside TechCrunch Sessions AI: Learn how Toyota and NLX use AI to parse millions of tech docs, with NLX CEO Andrei Papancea

    We’re down to the final two days before TechCrunch Sessions: AI unites the broader AI community at UC Berkeley’s Zellerbach Hall on Thursday, June 5! Expect a packed agenda with…

    Continue reading

    Leave a Reply

    Your email address will not be published. Required fields are marked *

    You Missed

    Yoshua Bengio launches LawZero, a nonprofit AI safety lab

    • By staff
    • June 3, 2025
    • 0 views

    Inside TechCrunch Sessions AI: Learn how Toyota and NLX use AI to parse millions of tech docs, with NLX CEO Andrei Papancea

    • By staff
    • June 3, 2025
    • 0 views

    DeepSeek may have used Google’s Gemini to train its latest model

    • By staff
    • June 3, 2025
    • 0 views

    Day 6 of 7: TechCrunch Sessions: AI Trivia Countdown — flex your brain, score 2-for-1 tickets

    • By staff
    • June 3, 2025
    • 2 views