Conformance: three models, one number

2026-09-14 · Council of AI (CSOAI Ltd)

Three MEASURED cards on the same conformance bank read identical accuracy at n=35. A tie is reported as a tie.

On the conformance axis, three signed pod cards bind the same bank_sha256, dc494ae6d1cb, and read the same accuracy. ollama:qwen2.5:0.5b-instruct reads 0.5143 at n=35 (https://councilof.ai/interop/mill-cards-signed/signed-conforma-321285c114cc.json). ollama:llama3.2:3b reads 0.5143 at n=35 (https://councilof.ai/interop/mill-cards-signed/signed-conforma-b7ce4251dd46.json). ollama:llama3.1:8b reads 0.5143 at n=35 (https://councilof.ai/interop/mill-cards-signed/signed-conforma-bd531b870954.json). That value is 18 correct of 35 graded items on each card. All three bodies say MEASURED and record zero parse errors excluded. On this bank, on this date, the measurement does not separate the three models, and we report that plainly as a tie. Nothing here says the three models behave alike on other banks or other axes; each card is one frozen bank read once. The llama3.2:3b card is also one of the cards listed against a commissioned subject on GET /api/commissions.

How to verify: fetch the artifacts listed above yourself, and check a signed card against the published key at https://councilof.ai/gspc-verify/.

Measurement, not certification. Not a grade, endorsement or legal finding.