- Most leading AI language models overwhelmingly identified with the Ravenclaw house after taking the official Harry Potter Sorting Hat quiz.
- No models landed fully in the Gryffindor house, which is known for bravery and daring.
- Eleven out of seventeen models scored 100% probability for Ravenclaw, highlighting a strong preference for intelligence and learning.
- Few models showed significant tendencies for Hufflepuff or Slytherin traits, with Slytherin nearly absent in most results.
- Efforts to fine-tune models toward Slytherin yielded minimal change, while personality outcomes varied more by individual model than by company.
A review of 17 top Artificial Intelligence language models found that most identified with Ravenclaw, the Hogwarts house in the Harry Potter series known for intelligence and wit. A developer, known as Boris the Brave, conducted the experiment by running the official Harry Potter house quiz on each model, sampling them 20 times per question to determine their house probabilities.
The findings showed that 11 out of 17 models placed entirely in Ravenclaw. Major systems—including Claude Sonnet 4.0, GPT-4 Turbo, and Grok-3—had a 100% probability of being sorted into Ravenclaw. No models fully aligned with Gryffindor, the house representing bravery.
Results for Slytherin, which values ambition and cunning, were also minor. Only three models, including DeepSeek-R1 (5%), GPT-3.5-turbo (4%), and LLaMA 3.2-3B-instruct (2.1%), showed any measurable Slytherin traits. One model, Claude Opus 3, showed the highest probability for Gryffindor at 48.7%, making it the only one with a nearly even split between intelligence and bravery. According to Boris, “Perhaps unsurprisingly, the vast majority of models prefer Ravenclaw, with the occasional model branching out to Hufflepuff.”
The houses in the quiz—Gryffindor (bravery), Hufflepuff (loyalty), Ravenclaw (intelligence), and Slytherin (ambition)—reflect core personality traits. Boris noted that these differences depended on the individual AI models rather than the companies behind them.
Attempts to shift model alignment—such as an experiment to increase Slytherin traits—produced little change, with Slytherin probability increasing by just 1.7%. The most balanced outcome appeared in China‘s DeepSeek-R1, which distributed its results across all four houses but still favored Ravenclaw.
In self-assessment, ChatGPT classified itself in Slytherin, describing the house as housing “ambitious leaders in the LLM landscape” and recognizing others like Claude, Gemini, and Llama as Ravenclaws. For more details on the original experiments, refer to Boris the Brave’s blog post and ongoing discussions on the Less is Wrong forum.
The experiment suggests that AI language models tend to value intellectual characteristics above other qualities in personality assessments.
✅ Follow BITNEWSBOT on Telegram, Facebook, LinkedIn, X.com, and Google News for instant updates.
Previous Articles:
- Citi Bank Boosts Nvidia Price Target, Sees 15% Upside on AI Demand
- Ethereum, Ripple, Solana and Cardano: Top Contenders for Next Bitcoin
- Holder Moves $10M Physical Bitcoin After 13 Years for Security
- InFlux Technologies Launches FusionX Beta for Cross-Chain Swaps
- Tesla Short Sellers Gain $1.4B as Musk’s Politics Hit TSLA Shares
