The autotrust/GEV-26B-Decide model, based on Gemma-4, has been released in two versions: a standard bf16 model and a quantized NVFP4 version optimized for vLLM. The NVFP4 version significantly reduces memory requirements, allowing it to fit on a 24GB GPU, and maintains performance comparable to the bf16 version on benchmarks like GPQA Diamond and HLE. The model is designed for decision-making tasks, with adaptive thinking capabilities that allow it to reason only when necessary, achieving fast inference times on hardware like the Nvidia B200. AI
IMPACT The NVFP4 quantization offers a path to running large models on more accessible hardware, potentially lowering the barrier for deploying advanced decision-making AI.
RANK_REASON Release of a specific model version with quantization details and performance metrics.
Read on Hugging Face Trending Models →
- autotrust/GEV-26B-Decide
- Gemma 4 26B A4B IT
- Hugging Face
- JEV-27B-VL
- MuJoCo
- Nvidia B200
- Nvidia RTX Pro 6000 Blackwell Workstation Edition
- transformers
- autotrust/GEV-26B-Decide-NVFP4
- Gemma-4
- Google Colab
- Kaggle
- NVFP4
- Nvidia
- RTX 5090
- vLLM
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →