Story

On air. The free wire runs 15 minutes behind live. See pricing

World, crypto, finance and tech news, scored for impact.

CryptoTokenPostKorean, translatedFiled 19 h ago

Heavy GPT-6 Astra test leads by 13.6%p in essential source type citations

Original headline필수 출처 유형 인용 13.6%p 앞선 헤비아 GPT-6 Astra 시험

Open the sourceReported by TokenPost, tokenpost.kr, 20 November 2026, by 서도윤

Not investment advice. Muhabbit scores how much a story may matter, not which way to trade.

OpenAI said its pre-release GPT-6 Astra model beat the next-best alternative by 13.6 percentage points in Hebbia's financial research test on required source-type citations, citing the results in a customer case study published on the 20th. In the same evaluation, 99.5% of cited excerpts were judged relevant to the question, a measure of citation relevance rather than overall answer accuracy or factuality. GPT-6 Astra also met 97.9% of applicable requirements in a presentation assessment, 14.4 percentage points above the strongest alternative. Hebbia's Matrix product builds answers from financial documents and internal materials and lets users trace the underlying evidence. The results matter as a signal of how frontier models perform on citation-grounded financial research, and the next thing to watch is whether the gains hold in broader, independent testing beyond a single vendor case study.

Muhabbit's AI wrote this summary in its own words from TokenPost's report and translated it from Korean. It can get details wrong, so check the source.

Named in this story

GPT-6 Astra, Hebbia, OpenAI