Ihteshamstar/qwen3-4b-cuad-extractor

TEXT GENERATIONPricing:Input $0.4 / Cached $0.08 / Output $0.8Concurrent Unit Cost:1Model Size:4BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Sep 13, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Cold

Ihteshamstar/qwen3-4b-cuad-extractor is a 4-billion parameter Qwen3-Instruct model fine-tuned with QLoRA for extracting 41 CUAD clause types from legal contracts. It achieves a detection F1 score of 0.900 and a strict F1 of 0.678, outperforming larger zero-shot models like Qwen3-14B and GPT-4.1 on this specific task. Optimized for single-pass extraction, it quotes contract text verbatim and explicitly returns empty lists for absent clauses, running efficiently on consumer hardware with ~2.5 GB (q4_K_M GGUF) memory.

Loading preview...

Overview

Ihteshamstar/qwen3-4b-cuad-extractor is a specialized 4-billion parameter Qwen3-Instruct model, fine-tuned using QLoRA to extract all 41 CUAD (Contract Understanding Atticus Dataset) clause types from legal contracts in a single pass. This model is designed to quote contract text verbatim and return explicit empty lists for clauses not present, making it highly efficient for automated contract review.

Key Capabilities

  • High Accuracy Clause Extraction: Achieves a detection F1 score of 0.900 and a strict full-span coverage F1 of 0.678 on the CUAD test set.
  • Performance Superiority: Outperforms zero-shot Qwen3-14B (0.816 F1) and even GPT-4.1 (0.641 strict F1) on the CUAD task, despite being significantly smaller.
  • Cost-Effective & Fast: Delivers comparable or better accuracy than Claude Opus 5.5 zero-shot at roughly 1/50th of the cost (~$0.6-1.4 per 1,000 contracts vs. ~$76) and processes contracts in 4.5 seconds per user.
  • Resource Efficient: Runs in approximately 2.5 GB (q4_K_M GGUF) on consumer hardware, making it suitable for local deployment.
  • Single-Pass Extraction: Extracts all 41 clause types simultaneously, reducing API calls and processing time compared to querying one clause type at a time.

Good For

  • Automated Legal Contract Review: Ideal for rapidly identifying and extracting specific clauses from English-language commercial contracts, particularly those similar to SEC EDGAR filings.
  • Reducing Manual Review Time: Significantly speeds up the initial phase of contract analysis by providing accurate, verbatim clause extractions.
  • Cost-Optimized AI Solutions: Offers a highly efficient and accurate solution for contract analysis at a fraction of the cost of larger, general-purpose LLMs.
  • Local Deployment: Suitable for deployment on consumer-grade GPUs using GGUF quantizations (e.g., Ollama) or with vLLM for higher throughput in production environments.

Finetunes
1 models