dragon-mistral-0.3-onnx

dragon-mistral-0.3-onnx is an ONNX int4 quantized version of dragon-mistral-0.3, providing a fast inference implementation, optimized for AI PCs using Intel GPU, CPU and NPU.

dragon-mistral-0.3 is a very accurate, new fact-based question-answering model, optimized for complex business documents, based on the most recent Mistral open source release v0.3.

Model Description

• Developed by: llmware
• Model type: mistral-0.3
• Parameters: 7 billion
• Model Parent: llmware/dragon-mistral-0.3
• Language(s) (NLP): English
• License: Apache 2.0
• Uses: Fact-based question-answering
• RAG Benchmark Accuracy Score: 99.5
• Quantization: int4

Model Card Contact

llmware on github

llmware on hf

llmware website