Hassabis amplifies claim that Gemini beats Claude and GPT with 'completely different' thinking

demishassabis · x · 2026-11-20

Demis Hassabis retweeted a viral thread by AlexFinn arguing that the new Google Gemini models are currently the best AI models available, above even Fable and Astra. The author's core claim: Claude and OpenAI models, while excellent, give roughly the same answers in different styles — as if trained on the same data and running the same race — while Gemini consistently offers novel, structurally different approaches, as if trained on a completely different dataset. He cites asking models for life advice, where ChatGPT and Claude felt sanitized. Individual impressions rather than benchmarks, but notable given Hassabis amplified it.

Related event: Hassabis Amplifies Praise for Gemini as Today's Best Model(2 posts)→

Original post →

More from Models

Models channel →