Upstage is an LLM and document AI platform from Seoul: the Solar model API, Document Parse and Studio agents on one account, paid by usage or per seat.
Upstage gives companies one place to read their documents and reason over what they contain. Upstage Co., Ltd., a Seoul AI company led by chief executive Sunghun Kim, builds and runs it. Every part sits on one account, so credits bought for the API also pay for work in Studio. The platform is aimed at document-heavy work in insurance, finance, healthcare and manufacturing. Solar Pro 4 is the flagship language model, with tool calling, strict JSON schema responses and a reasoning setting that trades speed for harder thinking. Solar Mini 4 takes high-frequency, cost-sensitive requests, Syn Pro serves Japanese business work, and Embed 2 turns text into vectors for search. Document Parse converts PDFs, scans, spreadsheets and slides into HTML or Markdown that a model can read. Information Extract pulls key-value data from invoices, claims and contracts without templates, while Document OCR and Document Classify handle plain text recognition and sorting. Upstage Studio runs agents that check invoices against contracts, stop duplicate claims and assemble underwriting files. Developers building document pipelines, retrieval systems or agents on top of a model are the Console's audience. Operations teams in claims, lending, KYC, revenue cycle and accounts payable use Studio when they would rather run agents than write code. The API is paid by usage, per token or per page, and teams that prepay a monthly or yearly commitment receive bonus credits, higher rate limits and faster support. Studio is sold per seat, starting with a no-cost plan, and larger organisations sign an enterprise contract with custom limits, connectors and on-premises deployment. A typical deployment parses a stack of loan or claim files with Document Parse, extracts the fields a reviewer needs and asks Solar to reason over the result. Teams on AWS can buy it through AWS Marketplace and run it in Amazon SageMaker, and developers can reach it from OpenAI Codex CLI or through an MCP server. Data that must stay in Korea can go to a Korean regional endpoint, currently in closed beta. Schools, hospitals and nonprofits can apply for no-cost access to Solar Pro and Document Parse. Upstage fits organisations whose decisions depend on documents being read in full and checked against each other.
Pricing
View plans-
Studio Free
Free
- No payment required; daily usage allowance that resets
- Email support
- 7-day trial of every Studio feature at a higher plan's usage level after signup
-
Studio Small
$20 per month
Per seat per month, billed monthly, excluding 10% VAT; no annual price shown
- More usage than Free, about 2,600 estimated pages a month per seat
- Auto-reload of usage
-
Studio Large
$200 per month
Per seat per month, billed monthly, excluding 10% VAT; no annual price shown
- More than 10 times the usage of Small, about 30,000 estimated pages a month per seat
- Auto-reload of usage
-
Studio Enterprise and Developer
Custom pricing
- Quote only
- Custom usage limits, SLA and role management
- Custom connectors
- On-premises or air-gapped environments by contract
Features
- Included: Published context window size
- Not included: Image input support
- Not included: Audio input support
- Included: Function and tool calling
- Included: JSON schema output mode
- Included: Extended reasoning mode
- Included: Prompt caching discount
- Not included: Batch inference API
- Not included: Fine-tuning offered
- Included: Open weights release
- Not included: Zero data retention option
- Included: Regional data residency
- Not included: Published safety evaluations
- Included: Free tier or trial credits
- Included: Enterprise agreement available
- Included: Published deprecation policy
- Not included: Published benchmark results
- Included: Embeddings endpoint
Additional Features
- Solar Pro 4 language model with a 512K context length
- Solar Mini 4 for high-frequency, cost-sensitive requests
- Syn Pro Japanese-focused business LLM
- Embed 2 passage and query embedding models
- Document Parse output to HTML and Markdown
- Information Extract key-value extraction without templates
- Document OCR text recognition
- Document Classify sorting by document type
Best for
- Insurers automating claims, underwriting and policy files
- Banks and lenders running KYC, credit and reconciliation work
- Hospital revenue cycle teams checking records against claims
- Manufacturers auditing invoices, contracts and shipments against terms
- Developers building document pipelines for retrieval and AI agents
- Operations staff who want document agents without writing code
- Teams needing a Japanese-focused model or Korean data processing
Use cases
- Parsing invoices into structured text
- Extracting vendor, amount and date from receipts
- Sorting mixed documents by type before extraction
- Checking freight invoices against contract rates
- Matching purchase orders to invoices
- Stopping duplicate insurance claims before payment
- Assembling a complete underwriting file
- Finding documentation gaps behind medical billing denials
- Moving KYC reviews through faster
- Converting legacy PDFs into data for AI pipelines
- Answering questions across long policy documents
- Building tool-calling agents on Solar Pro 4
Screenshots & Videos
Explore Upstage in action
User Reviews
Compare with Similar Software
Click to compare side by side with Upstage