The two public models show unusually high downloads, while the only structured score is an unverified author-run result.
AutoTrust decision models surge on Hugging Face without independent validation
2026-10-02Frontier watchIGNORE
AutoTrust published two Apache-2.0 decision-oriented models: GEV-26B-Decide, based on Gemma 4, and JEV-27B-VL, based on Qwen3.8. Their Hugging Face pages showed roughly 447,000 and 1.28 million downloads during this review. The only listed evaluation is an author-run, unverified Decision Index result, and JEV lists no model-index result.
Official · ReferenceAutoTrust GEV-26B-Decide model cardOfficial · ReferenceAutoTrust JEV-27B-VL model card
Availability: public weights and adapters
What this means for you
High download velocity is useful for discovery but does not establish model quality. Teams should not spend evaluation time until AutoTrust publishes a reproducible benchmark package or independent results appear. The current signal is unusually strong attention paired with weak verification.
FRONTIER WATCH
Public artifacts and validated results are different.
The recorded fact is the public work and its artifacts. Author-reported outcomes do not establish independent reproduction. Next checks include usable code, task conditions and third-party validation.
Evidence & related material
Technical details & scores
| Models | GEV-26B-Decide and JEV-27B-VL |
|---|---|
| Base Models | google/gemma-4-26B-A4B-it and Qwen/Qwen3.8-27B |
| License | Apache-2.0 |
| Downloads Checked 2026 10 05 | GEV 446,527; JEV 1,278,569 |
| Evaluation | GEV reports 62.48 on author-run Decision Index 0.2.1; result unverified; JEV has no model-index result |
| Availability | public weights and adapters |
Technical impact 2
Engineering value 1
Adoption signal 5
Maturity 2
Verification cost 4
Risk 3