models · qwen

Qwen3.8-27B-FP8

image-text-to-text model harvested from Hugging Face (trending). Catalogued at Quarantine — unverified, pending attestation.

Compare Verify

What this registry has checked

Quarantine. We hold the publisher’s own published hash for these weights and have recorded it against this entry. We have not downloaded and hashed the weights ourselves, so the hash is the publisher’s claim, not our measurement.
Methodhf-published-sha256

What is recorded

Publisherqwen
OriginRecorded as a covered nation under 10 U.S.C. § 4872(f) — the PRC, Russia, Iran or North Korea. The registry did not record which.
Licenceapache-2.0
Parameters27,781,427,952 (27.8B)
Native precisionF8_E4M3
Pulls recorded here123,157
Downloads reported upstream7,752,562

Measured on real hardware

None of these runs are ours. Each figure is the reporter’s, linked to their own evidence, under their own trust label. The number shown is single-stream decode — one user, concurrency 1 — because aggregate throughput across many users is a different quantity and reads as far faster than anyone will see.

Measured by local.ai

HardwareBuild measuredDecode @ 8KMemory @ 8KDecode mode
NVIDIA RTX PRO 6000 Blackwell 96GBfp8117 tok/s95.6 GBmtp:4
NVIDIA RTX PRO 6000 Blackwell 96GBfp8105 tok/s95.6 GBmtp
NVIDIA RTX 6000 Ada 48GBfp864.3 tok/s47.0 GBmtp:4
NVIDIA RTX 6000 Ada 48GBfp863.7 tok/s47.0 GBmtp:4
NVIDIA RTX 6000 Ada 48GBfp858.0 tok/s47.0 GBmtp
NVIDIA DGX Spark 128GBfp817.7 tok/s111.2 GBmtp:4
NVIDIA DGX Spark 128GBfp817.5 tok/s111.3 GBmtp:4
NVIDIA DGX Spark 128GBfp817.0 tok/s111.2 GBmtp

Measured by local.ai, not by us: 8 runs. Each names its engine pinned by image digest and the exact serve command, in the full record.

Elsewhere in the registry

Registry record as of 2026-09-01. Identity, licence and parameter count are properties of a release and do not move; tier can, and the verify link above re-checks it against the live registry rather than this page.