Cohere Parse Beats Frontier LLMs at Receipt Parsing

Melvin Vivas · X post · 2026-08-28 · Open on X

Topics: LLM Fundamentals, Evaluation (Evals) & Testing · Level: intermediate

Summary

The creator tested Cohere Parse on a hard receipt with many sub-items. It parsed the receipt with 100% accuracy, which no model had done for him before. He says it has the best price-performance for this use case, ahead of GPT 5.5, Gemini 3.5 Flash and Opus 4.8, and you can try it on a Hugging Face Space.

Key points

• Receipts with nested sub-items are a hard test for document parsing.
• Cohere Parse parsed his test receipt with 100% accuracy.
• For this use case, he says it has better price-performance than GPT 5.5, Gemini 3.5 Flash and Opus 4.8.
• A specialised parsing model can beat general frontier LLMs on cost and accuracy for document extraction.
• Test parsing models on your own hard documents before choosing one.

Resources mentioned

• Cohere Parse (Hugging Face Space) · tool · huggingface.co · free
  Cohere's document-parsing model, with a demo you can try on Hugging Face Spaces.
• Cohere · person · x.com · free · recommended by both Bashiri Smith & Melvin Vivas
  AI company that builds language, embedding and reranking models, and now the Transcribe speech-recognition model.
  Also in: Basic RAG Pipeline in 60 Seconds: From Documents to Grounded Answers (Bashiri Smith on Facebook · notes), Cohere Parse: Pricing vs Parse Bench Score (Melvin Vivas on X · notes), Cohere's North Micro Vision: a small open-source vision model for documents (Melvin Vivas on X · notes), Local Audio Transcription with Cohere Transcribe on WebGPU (Melvin Vivas on X · notes) and 1 more

Try this

• Try Cohere Parse on the Hugging Face Space with your own receipts or documents.
• Build a receipt-parsing pipeline that extracts line items and sub-items, and compare Cohere Parse with frontier LLMs on accuracy and cost.