Atlas

Benchmarks

← All benchmarks

FrontierFinance v3.1 — Factual Data Extraction Rubrics

Professional Work · 2026-07-06

Factual Data Extraction rubric-category slice in the official FrontierFinance v3.1 system breakdown, with a reported denominator of 8,547 rubric criteria. It reports micro-averaged rubric qualification rate within that displayed category.

Top models (higher is better)

ModelScore
Claude Fable 543.1
GPT-5.538.3
Opus 4.838.3
GPT-5.6 Sol35.7
Kimi K2.630.6
DeepSeek-V4-Pro30.2
GLM-5.228.7
Gemini 3.1 Pro Preview20.2
Loading Atlas data…