Ant Group releases FinFIRST, an open benchmark for financial search agents with 12,300 rubric points
FellMentKE · x · 2026-09-04
Ant Group, with professional support from CICC's investment banking team, released FinFIRST, an open benchmark for financial search agents. Version 1 includes:
- 123 expert-authored tasks
- 701 atomic criteria
- 12,300 rubric points
It ships alongside Ling-3.0-flash-Fin, an open model built for financial workflows that can connect to search, Python, databases, and spreadsheets — giving teams a starting point to test their own agent workflow ideas.
Related event: Ant Group Open-Sources FinFIRST Benchmark for Financial Search Agents(2 posts)→
More from Models
- More GPT-6 codenames spotted: GPT-6-Pro and mysterious 'WM' suffix fuel world model speculation — scaling01 · 2026-09-04
- Meta's speech model targets messy real rooms as the ears for its glasses and agent stack — heypearlai · 2026-09-04
- Meta claims top spot on Artificial Analysis for streaming speech-to-text and diarization at launch — heypearlai · 2026-09-04
- OpenAI ships GPT-6 Astra; Brockman suggests it could be AGI — steph_palazzolo · 2026-09-04
- Follow-up links added to GPT-6 Astra leak thread — Angaisb_ · 2026-09-04
- GPT-6-Astra, GPT-6-Pro codenames spotted in ChatGPT frontend, fueling unconfirmed rumors — Angaisb_ · 2026-09-04