CCooScope.

Explainer

Open weights, open code, open enough for what?

Inspect the actual artifacts and permissions behind an openness claim instead of treating open as a single product feature.

4 min estimated readStart hereEvidence baseline

THE SHORT ANSWER

The useful question is what you can inspect, modify, run and redistribute in the task you have.

Why it matters

Downloading weights may help you run a model, but it does not tell you how the model was trained or which parts you can change. For a research project, a training recipe may matter most. For a deployed tool, reproducible inference and a clear migration path may matter more.

Read the evidence

The Open Source Initiative’s AI Definition 1.0 treats data information, code and parameters as distinct requirements. Ai2’s 26 November 2024 OLMo 2 announcement lists weights, data, code, recipes and intermediate checkpoints. Use that inventory as an example of what to look for, not as a claim that one model is best for every use.

How to use this resource

Build an artifact matrix with one row per component. Make unknowns visible instead of filling them with the word open.

  • List the weights, tokenizer, inference code, training code and data information separately.
  • Record the license attached to each artifact and any linked use conditions.
  • Ask what would be needed to reproduce, adapt or replace this component.
  • Check model and dataset terms separately from the software library’s license.

Limits and evidence

This is a technical reading checklist, not legal clearance or a compliance certification. The OLMo example is historical. No training run or reproducibility audit was performed, and an available artifact does not guarantee sufficient hardware, documentation or task quality.

Put it to work.

  • Separate model, code and data.
  • Keep an artifact-level license record.
  • Record missing reproduction inputs.
  • Check the intended use before adopting.

Go to the source.

Original material, with its own scope and publication date. This note is an explanation, not a CooScope performance test.

  1. Open Source Initiative — Open Source AI Definition 1.0Link / source review: 2026-10-04
  2. Ai2 — OLMo 2 release and artifact inventory (26 November 2024)Link / source review: 2026-10-04
Evidence baseline and source-review dates are different. A recent link check does not make an older study a current product test. Read our method →

Keep the thread going.

Explore this topic →