GPAI Ledger › GPAI Training Transparency tracker (AI Accountability Lab (AIAL)) › Capture 12 Sep 2026
Velvet 25B — capture 20260912T063135Z
Filed under AI Accountability Lab (AIAL) — GPAI Training Transparency tracker, the source this project captured it from; the assessment itself is of the model named above.
| Provider | Almawave |
|---|---|
| Target | AIAL evaluation (earlier state) — https://raw.githubusercontent.com/AIAccountabilityLab/gpai-training-transparency/40ed45d5cdd76319e1ea94d0f42b2cb35eee7dfa/evals/velvet-25b.yaml |
| Fetched (UTC) | 2026-09-12T06:31:35Z |
| Upstream commit | 11 Sep 2026 — 40ed45d5cdd7 (when this state began to stand in the upstream repository; this archive fetched it at the time above, not then) |
| Stored file | 3,796 bytes — not served (third-party research, archived but not republished here); the hash and timestamp proof below still establish what was published when |
| SHA-256 | 1f2b5b6cccf362ca227c48b0db79c0d3efba3dd2ecc36b466097691ac5265382 |
| OpenTimestamps proof | 1f2b5b6cccf362ca227c48b0db79c0d3efba3dd2ecc36b466097691ac5265382.yaml.20260912T063135Z.ots (calendar-attested; anchored in bitcoin over time) |
| Wayback | not saved |
| Prior capture of this target | — first capture of this target |
Verify: obtain the file from the target address above, then sha256sum it — it must equal the hash above; the .ots proof dates that hash.
Structured facts
This is another organisation's own research, and this project does not redistribute it: the file itself is not served here; these facts still pin its identity and let any copy be authenticated against this record.
| Document size | 3,796 bytes |
|---|---|
| Canonical text SHA-256 | d640b753fd5f01f3845547898b53266524c5be9e737401dd2e74608f41553c66 |
| Extracted text length | 434 words |