Ling-3.0-flash-Fin trails VL in reported knowledge-work benchmark scores
ArtificialAnlys reports that Fin scored 967 Elo on AA-Briefcase versus VL’s 986, with a slightly higher presentation score but lower analytical quality.
TLDR
ArtificialAnlys reports that Ling-3.0-flash-Fin scored 1171 Elo on GDPval-AA v2, which tests agents on professional knowledge work. That puts it behind Ling-3.0-flash-VL at 1225 but ahead of MiniMax-M2.7 at 1087. On AA-Briefcase, which tests complex business workflows using large file collections to produce spreadsheets, presentations and memos, it reports Fin at 967 Elo versus VL at 986. Fin passed fewer rubric checks (23.5% vs. 24.9%) and scored lower on Analytical Quality Elo (866 vs. 907), but slightly higher on Presentation Elo (1095 vs. 1076), despite lacking VL’s image-input capabilities.
Combined views
186
1 Source, first seen 15d ago