MrRoyaleAce/nyaya-7b

TEXT GENERATIONConcurrency Cost:1Model Size:7BQuant:FP8Ctx Length:4kTool Calling:SupportedPublished:Jun 12, 2026License:apache-2.0Architecture:Transformer Open Weights Cold

MrRoyaleAce/nyaya-7b is a 7 billion parameter, domain-adapted instruction-finetuned version of Mistral-7B-Instruct-v0.3, specifically designed for parsing Indian Supreme Court and High Court judgments. It excels at extracting structured legal information into validated JSON, offering a zero-cost, offline solution for legal NLP tasks. This model is optimized for accuracy in legal data extraction, outperforming commercial alternatives like Gemini 2.5 Flash in key metrics such as statute F1 score and outcome accuracy.

Loading preview...

Nyaya-7B: Indian Legal Judgment Parser

Nyaya-7B is a specialized 7-billion parameter language model, fine-tuned from Mistral-7B-Instruct-v0.3 using QLoRA (4-bit NF4 quantization). Its primary function is to extract structured legal information from raw Indian Supreme Court and High Court judgments, converting them into clean, validated JSON format. This model is designed for offline use, providing a zero-cost solution for legal data extraction.

Key Capabilities

  • Structured Data Extraction: Parses judgment text to extract fields like case_name, citation, court, year, petitioner, respondent, subject_matter, statutes_cited, precedents_cited, legal_issues, holding, and outcome.
  • High Accuracy: Benchmarked against Gemini 2.5 Flash on a 50-case test set, Nyaya-7B achieved a Statute F1 score of 0.425 (87% higher) and Outcome Accuracy of 0.64 (220% higher). It also demonstrated a 45% lower hallucination rate.
  • Offline & Cost-Free: Operates entirely offline, eliminating API costs associated with commercial models.
  • Domain-Adapted: Trained on over 10,000 Indian Supreme Court and High Court judgments, curated from datasets like ILSum and InLegalNLP.

When to Use Nyaya-7B

  • Legal Research & Document Processing: Automate the extraction of key details from Indian legal documents.
  • Paralegal Workflow Tools: Integrate into tools for legal analytics and case record digitization.
  • Academic Research: Utilize for studies in Indian legal NLP.
  • Building Legal Knowledge Systems: Develop search engines or knowledge graphs based on structured legal data.

Limitations: Not intended for legal advice, may perform poorly on pre-1950 judgments or non-English text, and requires OCR preprocessing for scanned documents. Critical citations should always be validated.

Finetunes
1 models