본문 바로가기
Chroma's New 20B Model Beats GPT-5 at Search

Chroma's New 20B Model Beats GPT-5 at Search

AIPrompt Engineering· 2026-03-27

Chroma AI has released Context-1, a 20 billion parameter self-editing search agent. This model is specifically trained for retrieval augmented generation (RAG) tasks, outperforming larger models in retrieval benchmarks at a lower cost and higher speed, and offers a new approach…

핵심 요약

  • Context-1 is a specialized large language model (LLM) trained for retrieval augmented generation (RAG), based on the GPT-OSS 20 billion parameter model.
  • Traditional RAG systems suffer from losing global context, inability to cross-reference documents, and semantic similarity not always equating to relevance.
  • Agentic RAG involves multiple retrieval steps, where an LLM plans and executes searches, allowing for iterative information gathering and plan updates.

전체 요약과 종목별 의견·실시간 분석을 보려면 로그인하세요.

로그인 / 회원가입