Krasnopjorovs/gemma-4-26b-a4b-it-Imatrix-IQ4_XS-GGUF - Каталог нейросетей
Генерация текста

Krasnopjorovs/gemma-4-26b-a4b-it-Imatrix-IQ4_XS-GGUF

Добавлено:
Krasnopjorovs/gemma-4-26b-a4b-it-Imatrix-IQ4_XS-GGUF

This is a high-performance, hybrid Mixture of Experts (MoE) GGUF version of Google’s Gemma 4 26B. By utilizing a custom Importance Matrix (imatrix), this IQ4XS` quantization maintains extreme logical precision while enabling blazing fast inference speeds. In this MoE architecture, only a fraction of parameters are activated per token, making it significantly faster than dense models of similar size. The model retains the full 262,144 token vocabulary, showing exceptional reasoning in English, Russian, German, and other major languages. Base Model: Google Gemma 4 26B-A4B (MoE) Quantization Format: IQ4XS (GGUF) Optimization: Custom Imatrix applied File Size: ~13.92 GB (Extremely VRAM efficient) Context Size: 131,072 tokens License:** Gemma The model was subjected to rigorous reasoning and programming benchmarks: 1. Coding (Autograd from scratch): Successfully wrote a complete Value class for automatic differentiation in Python (Scalar Autograd) with topological sort and backpropagation logic. 2. Physical Reasoning: Correctly predicted the gravitational outcome of an overturned table with an open water bottle (zero hallucinations). 3. Lateral Thinking: Solved the «Snail on the…

Модальности:
Генерация текста

Области применения:
Диалог / чат


Задача: Генерация текста
Автор: Krasnopjorovs
Теги: gguf, gemma, gemma-4, imatrix, quantization, moe, llama-cpp, en
Лайков: 4  |  Загрузок: 601

Открыть на HuggingFace →

Описание основано на материалах HuggingFace. Перевод выполнен автоматически.