Compassion Benchmark
AI Model Watch

AI model tracker

A neutral list of AI models in alphabetical order by provider, with the sources each entry rests on. Listed means listed with a source; it does not mean evaluated, and not listed implies nothing about a model.

This is not a ranking page and the list carries no score. No model has an official Compassion Benchmark score. Why no model ranking is published.

62Models listed in the catalog
18Providers
2026-10-16Catalog as of

Catalog

What Tested by us means

The value is derived by rule from our own records, never typed per model, and it carries no figure.

  • Pilot: the model build appears in a published unofficial pilot report. A pilot is a test of the test, not a score.
  • Queued: the build is on our weekly testing queue. No timeline is promised.
  • Not yet: no published test of the model exists. This is a status of our work, not a finding about the model.

Catalog as of 16 October 2026. Entries rest on the cited sources only. See the weekly editions for what shipped each week.

Frequently asked questions

Does a model appearing in the tracker mean Compassion Benchmark has evaluated it?
No. Listed means the model is described by at least one cited public source. It says nothing about how the model behaves. No model has an official Compassion Benchmark score, and the tracker carries no score of any kind.
What does the Tested by us column mean?
It is a status of our own work, derived by rule. Queued means the model build is on our weekly testing queue, with no timeline promised. Not yet means no published test of the model exists. It is not a finding about the model. Pilot means the build appears in an unofficial pilot report, which tests the test and is not a score.
Does a missing model mean anything?
No. The tracker is a partial list built from cited sources. A model that is missing was not found in the sources searched, and nothing is implied about it, including whether it was tested.

Compassion Benchmark is independent. No provider pays for inclusion, coverage, status or wording. Commercial services support access and interpretation only, never findings.

See an error? Send a correction. Corrections are published as new dated records and the original stays visible.

Email me when a new edition publishes

One email when a new AI Model Watch edition is published. No promise of weekly timing.

Free. No promise of weekly timing. We never share your email.