Upstage

What Upstage does

Upstage builds high-performance AI technology, including industry-leading document processing engines and large-language models (LLMs), to boost work efficiency. Their flagship LLM, Solar, is optimized for speed and groundedness, providing cost-effective and powerful AI solutions for enterprises. Their Document AI, featuring advanced AI OCR technology, automates document processing with high accuracy. Upstage's solutions are designed for industries where accuracy and reliability are critical, offering flexible deployment options including SaaS, AWS Marketplace, and on-premise.

Products

• Solar LLMAn enterprise-grade language model family optimized for speed and groundedness, available in versions like Solar Pro 2, Solar Pro 3 (102B parameters), and Solar Mini. It defines the benchmark for cost-effective and powerful AI solutions for enterprises.
• Document ParseConverts PDFs, scans, and emails into clean, machine-readable text (HTML, Markdown) ready for AI pipelines. It features a vision-language model for high accuracy on complex layouts, tables, charts, and handwritten elements, and offers fast processing speeds.
• Information ExtractPulls structured key-value data from invoices, claims, and contracts with audited accuracy. It is schema-agnostic, adaptable, and understands context and intent, working with any document type without requiring training or templates.
• Upstage StudioA platform for teams to design, deploy, and operate AI-powered workflows across document-heavy processes. It allows building agents by connecting AI nodes, using pre-built templates, and offers features for review, optimization, and real-time monitoring of accuracy, speed, and stability.
• Upstage OCRAdvanced AI OCR technology for automating documents with high precision, achieving over 95% accuracy across diverse document types, including handwritten notes and complex forms.

Key capabilities

• High-performance AI built by experts
• Industry-leading document processing engines
• Large-language models (LLMs)
• Advanced AI OCR technology
• Optimized for speed and groundedness (Solar LLM)
• Converts PDFs, scans, and emails into clean, machine-readable text (Document Parse)
• Pulls structured key-value data from invoices, claims, and contracts (Information Extract)
• Enterprise-grade AI models
• Flexible deployment: SaaS, AWS Marketplace, on-premise, hybrid
• Vision-language model (VLM) for handwriting, tables, charts, stamps, and complex layouts
• Built-in verification and consistency checks for extracted data
• PHI protection and HIPAA-ready features
• Automatic detection and redaction of PHI/PII
• Summarization capabilities for clinical records
• Schema-agnostic and adaptable extraction (Information Extract)
• Seamless integration via API with existing tools (ERP, CRM, core banking systems)
• Real-time monitoring and audit logs
• Controllable data retention
• SSO/Directory Integration
• Guardrail for AI responses
• Enterprise Security: SOC 2 Type 1, ISO 27001/27701 certified, HIPAA compliant
• Role-Based Access Control (RBAC)
• Full Audit Logs & Traceability
• Fast processing speed (Document Parse: 0.6 seconds per page on average)
• Unmatched accuracy (Document Parse: 5%+ higher layout & table recognition vs. document processing models)
• Zero training for Information Extract
• Handles hundreds of pages at once (Information Extract)
• Rebuilds tables across page breaks (Information Extract)
• Understands deeply layered layouts (Information Extract)
• Corrects orientation automatically (Information Extract)

Use cases

• Automate claims, underwriting, and policy workflows in insurance
• Automate medical record and operational document work in healthcare
• Connect healthcare data to EMR and hospital systems
• Automate document work across lending, KYC, and risk in financial services
• Automate document work across quality, supply chain, and compliance in manufacturing
• Convert complex documents into LLM-friendly formats for precise and context-aware search (RAG)
• Improve knowledge retrieval by making complex documents fully searchable and accessible
• Automatically generate concise and informative summaries from lengthy reports, contracts, or research papers
• Digitize legal documents, contracts, and compliance reports into clean, digital formats
• Make regulatory filings and agreements easier to search, analyze, and review
• Preserve tables, charts, and text layouts when converting financial documents into digital formats
• Improve accessibility and readability of complex reports by converting them into HTML or Markdown
• Automate intake, extraction, routing, and risk-flagging for claims and underwriting pipelines
• Extract obligations, exclusions, and compliance risks across large document sets for contract and policy review
• Extract and validate key financial terms from diverse invoice and PO formats
• Automate group benefits underwriting
• Claims payment review automation
• Submission packet processing
• FNOL (First Notice of Loss) accident notice intake
• Loss runs and declaration forms standardization
• Clinical record extraction & summary
• Paper chart to EHR digitization
• Clinical trial operations (data collection and extraction)
• PHI de-identification for research
• Procurement & AP automation in manufacturing
• Insights from production & maintenance records
• Semantic drawing & record search in manufacturing
• Instant SOP & manual search for operators
• Onboarding & KYC automation in financial services
• Unstructured contract data extraction in financial services
• Audit & regulatory reporting in financial services
• Repetitive back-office automation in financial services
• News translation pipeline
• Streamlining data management
• Simplifying tax deferral registration
• Automating RFQ email processing
• Transforming insurance claims processing
• Automating underwriting with OCR and Solar LLM
• Real-time life insurance claims digitization
• Knowledge retrieval and report automation
• Automating product price comparisons
• AI-driven legal automation for faster case handling
• Transforming news search with RAG

AI approach

Upstage develops high-performance AI solutions, including large language models (LLMs) like Solar and Document AI with advanced AI OCR technology. Their technology focuses on document processing, information extraction, and understanding complex, messy documents across various industries. They emphasize accuracy, reliability, and enterprise-grade security.

Tech named: large-language models (LLMs), Document AI, AI OCR technology, Solar LLM, Document Parse, Information Extract, Vision-Language Model (VLM), MoE language model with 102B parameters (Solar Pro 3), Retrieval-Augmented Generation (RAG)

Industries served

• Insurance
• Healthcare
• Financial Services
• Manufacturing
• Government
• NPO & AWS Public Sector
• Media

What it says sets it apart

• High-performance AI built by experts, ensuring accuracy and reliability
• Proprietary large language model (Solar LLM) optimized for enterprise use
• Advanced Document AI with AI OCR technology for unmatched precision across document types
• Vision-language model (VLM) capable of reading handwriting, tables, charts, stamps, and complex layouts, understanding context and intent
• Flexible deployment options: SaaS, AWS Marketplace, Azure Marketplace, on-premise, air-gapped, VPC
• Built-in verification and consistency checks for extracted data, routing only low-confidence fields for human review
• PHI protection and HIPAA-ready with automatic detection and redaction of sensitive data
• Schema-agnostic and adaptable extraction that dynamically processes and generates structured outputs aligned to any given schema
• Seamless API-first integration with existing ERP, CRM, core banking, and MES systems without rip-and-replace
• Upstage Studio for designing, deploying, and operating AI-powered workflows with full visibility, control, and governance
• Faster processing speeds (e.g., Document Parse is 5-10x faster than competitors)
• Higher accuracy in layout and table recognition compared to other document processing models (DP-Bench benchmark)
• Zero training required for Information Extract, reducing setup time and maintenance
• Ability to handle hundreds of pages, rebuild tables across page breaks, and understand deeply layered layouts
• Certified enterprise protection: SOC 2 Type 1, ISO 27001/27701, HIPAA compliant
• Query-based integration for financial systems to avoid re-engineering core systems
• Proven track record with leading companies worldwide and global awards/industry rankings

Funding rounds we track

From the AI funding tracker — rounds as reported by the linked publications.