12ModelsAbout 3 minutes
How open is this AI?
“Open” can mean free access, downloadable weights, permission to reuse, or enough information to inspect a release. Eight questions help you tell those claims apart.
Start the eight questionsThe distinction test
Sources and definitions ↓Loading question…
Your result
This counts answers to eight teaching scenarios. It is not an openness score for any model or a test of a person's expertise.
Definitions and source trail
A claim about a release needs a precise boundary.
These questions distinguish what a person can access, what the terms permit, what a release discloses, and what a deployment requires. One company can publish releases with different terms. No answer assigns a permanent label to a company.
- Open Source Initiative, Open Source AI Definition 1.0 and FAQ: freedoms, parameters, training code and sufficiently detailed data information. OSI does not certify individual models.
- OpenAI, open models and GPT-6 Astra availability: separate downloadable and hosted releases from one organization.
- Meta, Llama 4 Community License: an example of release-specific conditions on use and redistribution.
- Hugging Face, Inference Endpoints pricing and NIST Privacy Framework: hosting costs and privacy controls depend on deployment choices.
Checked September 27, 2026. The quiz teaches general distinctions; the named releases illustrate them. It does not adjudicate a particular model's license, quality, security, or total cost. Terms and availability can change. Answers stay in this browser; only aggregate interaction counts are sent.
Continue with the argument behind the tools in Intelligence Capitalism.
Browse all tools →