Speed
Low latency and realtime
Answer before the user notices: voice agents, autocomplete and live UI, where the first token has to arrive in a few hundred milliseconds.
Job page with picks by constraint
What are you trying to do?
Pick the job, then the constraint you can't move: speed, the hardware you own, the size of the input, the license. Each page starts from today's data and says where every number came from.
Jobs, by group
Speed
Cheap batch extraction
Pull fields out of thousands of documents at the lowest cost per page.
Where it runs
On-device and laptop
Run it on a phone, a 16 GB laptop or a 64 GB Mac, with no server in the loop.
Job page with picks by constraint
Where it runs
Single-GPU self-host
Serve it yourself on one GPU with 24, 48 or 80 GB of memory.
Size of input
Search and RAG
Find the right passage first: embeddings to retrieve, rerankers to put the best on top.
Doing things
Coding agents
Write, fix and review code with an agent in the loop.
Job page with picks by constraint
Doing things
Tool-calling agents
Call functions and APIs reliably, from one tool call to a browser agent.
Seeing
OCR and forms to JSON
Turn scans, forms and tables into structured JSON.
Job page with picks by constraint
By industry
Thin pages that bundle the three or four jobs an industry hires models for, with its constraints (data residency, license gates, an audit trail) as filters. Drafts, under review.
- Legal Contracts, filings and discovery: long documents, scanned exhibits and strict rules about where the text goes.
- Healthcare Clinical notes, intake forms and claims: patient data that usually has to stay on hardware you control.
- Finance Statements, invoices and filings at volume, where cost per page and a reviewable trail both matter.
- Customer support Tickets, chats and calls: fast replies, the right tool call, and nothing unsafe reaching a customer.
- Software teams Coding agents, code review and internal tools, often with a rule that source code stays in-house.