What Compresr does
Compresr is a context compression service for large language models that reduces the number of tokens in a prompt while preserving the information needed to answer a query. It analyzes both the document and the specific question, then removes irrelevant tokens, lowering API cost and latency at the same time.
Key capabilities
The compression is question-aware, retaining only the tokens relevant to the user's query, and the company reports up to 90% token reduction on sparse documents with typical ratios around 10x. On its FinanceBench testing over SEC filings, Compresr reports maintaining or improving accuracy while cutting cost. It is available as a hosted SDK in TypeScript and Python or as an on-premises deployment, with free starter credits and per-million-token production pricing.
Who it's for
Compresr serves developers and teams running LLMs over long documents such as financial reports and PDFs who want lower token spend without sacrificing answer quality, including enterprise, finance and healthcare users who need on-premises deployment.