GEV-26B-Decide (AI model): news timeline & CVEs

autotrust/GEV-26B-Decide-NVFP4 — new model trending #30 on Hugging Face

autotrust published GEV-26B-Decide-NVFP4, an NVFP4 expert-quantized 26B model that fits in about 17 GB.

autotrust released GEV-26B-Decide-NVFP4, quantizing the 3,840 routed expert MLPs of GEV-26B-Decide to NVIDIA NVFP4 while leaving attention, routers, the vision tower, and decision heads in bf16. Download size falls from 49.5 GB to 18 GB, and vLLM weight memory from 51.1 GiB to 17.1 GiB. On System 1, GPQA Diamond moves from 43.9% to 44.9% and HLE text multiple choice from 8.4% to 9.7%, within run-to-run noise. The bf16 model scores 62.48 on Decision Index 0.2.1 with adaptive thinking.

Hugging Face trending modelsOct 5, 01:31 UTC · 4d agoModel release 2 sources2

autotrust/GEV-26B-Decide — new model trending #30 on Hugging Face

AutoTrust released GEV-26B-Decide, an open 26B Gemma-4 decision model scoring 62.48 on Decision Index 0.2.1.

AutoTrust AI published autotrust/GEV-26B-Decide, open weights previously named JEV-Gemma4-26B-A4B, built on Google's Gemma-4-26B-A4B-it (26 billion parameters, about 4 billion active). On Decision Index 0.2.1 it scores 62.48 balanced skill, above the TypeSafe Jev 1.13 board mark of 57.91; area scores include Tools and Automation 0.697 and Arts and Taste 0.415. System 1 returns calibrated option probabilities in one pass, about 45 ms on a B200, and adaptive thinking invokes Gemma-4 reasoning below 0.8 confidence. In demos it finished 95% of 60 browser tasks at about 85 ms per click, but only 40% of 20 robot-arm pick-and-place scenes.

Hugging Face trending modelsupdated Oct 5, 01:31 UTC · 4d agofirst Oct 2, 04:20 UTC · 6d agoModel release 2 sources1

autotrust/JEV-27B-VL — new model trending #30 on Hugging Face

autotrust published JEV-27B-VL, a vision model that makes calibrated System 1 decisions over images in one pass.

autotrust/JEV-27B-VL adds vision to JEV-27B. System 2 is the unmodified Qwen3.8-27B with optional step-by-step reasoning on images. System 1 answers yes/no, multi-choice, or 0–5 ratings over text and images and returns a calibrated probability for every option in one forward pass, with prompts up to 256K tokens. Reported demos completed 75% of 20 robot-arm pick-and-place scenes and 95% of 60 multi-step computer-use tasks, and cover-only ranking reached 0.727 AUC on MicroLens-100k.

Hugging Face trending modelsSep 30, 05:02 UTC · 8d agoModel release1

GEV-26B-Decide

Timeline3 articles

Appears with

autotrust/GEV-26B-Decide-NVFP4 — new model trending #30 on Hugging Face

autotrust/GEV-26B-Decide — new model trending #30 on Hugging Face

autotrust/JEV-27B-VL — new model trending #30 on Hugging Face