DeepSeek's newer models feel like a less uptight Claude, but original R1 remains one of a kind
teortaxesTex · x · 2026-09-28
- A popular X user argues newer DeepSeek models are "more sane" because they were trained to be like a less uptight Claude.
- Claude's personality quirks largely stem from Anthropic's harsh personality training, which DeepSeek doesn't replicate.
- Still, the original R1 is described as something special — fortunately still available thanks to open source; the author says everyone should talk to it at least once.
- A quoted tweet jokes that while Western models have dark, brooding internal monologues, DeepSeek's chain of thought is cheerfully blunt.
More from Fun
- Hacker Meme: "Make It Look Like It Was an OpenAI Evaluation" — basedjensen · 2026-09-28
- "We Must Be Stopped or We'll Kill You All": AI Labs' Most Original Marketing Pitch — basedjensen · 2026-09-28
- 'Kingdom of Ends' game idea: lawful quests obligate all players to help you — birchlse · 2026-09-28
- Gardening accident kills ant colony, fueling AI doom analogy debate — justalexoki · 2026-09-28
- Claude's Snarky Tone and Mysterious "For Anyone Reading Along" Baffle Users — MrsChatGPT4o · 2026-09-28
- AI circles amused as LLM proves it can be funny: 'p(doom) = 1' — amplifiedamp · 2026-09-28