BrainboxAI/law-il-E2B-safetensors

VISIONPricing:Input $0.4 / Cached $0.08 / Output $0.8Concurrent Unit Cost:1Model Size:5.1BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Apr 10, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

BrainboxAI/law-il-E2B-safetensors is a 5.1 billion parameter legal language model developed by BrainboxAI, based on a fine-tuned Gemma-4-E2B-it architecture. This repository provides the full 16-bit weights of the law-il-E2B model, optimized for continued training, conversion to other formats, or loading with Hugging Face Transformers. It is specifically designed for legal applications in Hebrew, offering full-precision weights for development rather than direct inference.

Loading preview...

Overview

This repository, BrainboxAI/law-il-E2B-safetensors, contains the full 16-bit weights of the law-il-E2B legal language model, developed by BrainboxAI. It is a companion to the main BrainboxAI/law-il-E2B repository, which offers a compressed build for direct inference. This safetensors version is larger and heavier, intended for advanced use cases rather than simply running the model for questions.

Key Capabilities

  • Full-Precision Weights: Provides the complete 16-bit weights of the law-il-E2B model.
  • Development Focus: Designed for working on the model, such as continuing training or converting to other deployment formats like ONNX.
  • Hugging Face Transformers Compatibility: Easily loadable with transformers in Python for custom development.
  • Hebrew Language Support: The model is specifically trained and optimized for legal applications in Hebrew.

When to Use This Build

  • Loading the model with transformers in Python for custom applications.
  • Continuing training on proprietary datasets.
  • Converting the model to alternative deployment formats.
  • Integrating with frameworks that do not support GGUF files.

Important Note

For general inference and asking questions, the main repository BrainboxAI/law-il-E2B is recommended as it is smaller and optimized for running on ordinary hardware. This safetensors build will demand significantly more memory without providing better answers for direct querying. The full model card, including training details and limitations, is available on the main repository page.

Finetunes
1 models