ai model report

Is cross-encoder/mmarco-mMiniLMv2-L12-H384-v1 safe to load?

CLEAN risk. 8 of 186 tracked dependencies carry advisories. Driven by No version-matched advisories and no malicious packages..

View on HuggingFace

// posture

fig. 01 — posture on eight axes, 100 is clean. Weakest: load safety at 60.

// the finding

not scoredno advisory applies to a version this model pinsCLEAN

LOWShips safetensors plus 1 PyTorch pickle weight file(s) (pytorch_model.bin) that Hugging Face loaders (transformers, diffusers, PEFT) load instead of the safetensors when a consumer passes use_safetensors=False; and 2 pickle checkpoint(s) for other tooling (openvino/openvino_model.bin, openvino/openvino_model_qint8_quantized.bin); from_pretrained never loads these, even with use_safetensors=False, but code that torch.load()s them directly (a reference implementation, a training script) executes pickle.

2 more findings on this model.

Author: cross-encoder
Task: text-ranking
Loader: sentence-transformers
License: apache-2.0
Downloads: 2,336,931
Remote code: not required

// dependencies

13 direct, 12 with findings, 177 reachable

Package	Evidence	Advisories	Worst
transformers	direct, runtime	24	none pinned
torch	direct, runtime	13	none pinned
numpy	direct, install	8	hidden
pytorch-lightning	direct, install	6	hidden
scikit-learn	direct, install	3	hidden
tqdm	direct, install	2	hidden
datasets	direct, install	1	hidden
sentence-transformers	direct, install	1	hidden

Create a free accountto see all 186 dependencies, which versions are pinned, and what to change.

Risk computed 2026-11-06 at 20:45 UTC. Dependencies come from the model's requirements, its declared loader and the code it ships; advisories from NVD, GHSA and OSV.