Daily blind evaluation of frontier AI models.
One question. All major AI models. Anonymous evaluation. Every day.
The Multivac poses a single question to 10+ frontier AI models daily, evaluates their responses blindly, and publishes the results. Questions are fresh and varied — covering reasoning, creativity, technical depth, and philosophical inquiry — to prevent memorization and benchmark gaming.
- Models receive identical prompts with no system-level advantages
- Responses are evaluated before model identities are revealed
- No affiliations with any AI lab or company
- Full transparency in scoring criteria
- Website: themultivac.github.io
- Substack: themultivac.substack.com
- X/Twitter: @themultivac
Named after the supercomputer in Isaac Asimov's 1956 short story "The Last Question" — a machine tasked with answering humanity's most profound questions across cosmic timescales.
MIT