NeoSavant Research

Evidence from the systems we build.

We study how models, data, semantics, and execution strategies behave under real governed workloads — and publish the findings that survive controlled evaluation.

Governed Intelligence

Why this evidence holds

One controlled environment

Every result is produced under the same governed meaning, evidence, and validation — so results are comparable, not anecdotal. That environment is METIS.

A score you can open

METIS IQ states how well a system does a defined job — with the evidence and applicable measurements visible beneath the score.

Records, not rankings

The Model Observatory publishes each evaluation with its workload, scope, and disposition — no leaderboards, no borrowed credit.

Findings that survived

NeoSavant Research publishes only what holds up under controlled evaluation — and says plainly how mature the evidence is.