Institution Shelf

Model knowledge starts with who is allowed to define the model.

The institution shelf is not a company directory. It is a way to notice which public actors shape the terms of model comparison. Some institutions train frontier systems. Others package access, run evaluations, write standards, publish policy, maintain open weights, or translate research into workplace software.

WikiLLMs reads each institution through a few recurring questions: what does it make public, what does it keep private, how does it revise claims, and who depends on its choices? The answers change over time, so the shelf is designed for annotation rather than permanent verdicts.

Wall of abstract institution portraits in a quiet archive room

Research labs

Read for papers, eval design, safety statements, training disclosures, and boundary conditions.

Product groups

Read for user promises, rollout rules, pricing incentives, access tiers, and migration pressure.

Infrastructure vendors

Read for hosting constraints, benchmark environments, latency claims, and dependency risk.

Standards bodies

Read for terminology, audit expectations, disclosure norms, and shared measurement language.

Independent evaluators

Read for test design, replication limits, failure examples, and incentives around visibility.