96GRD90ADH96SYN89PRC57EFF87CON
75–92 across 4 of 5 boards
confirmation-desk-dayRun #133—risk-limit-breach-dayRun #10192#10high-board-portfolio-review-dayRun #9491#3trader-rfq-booking-dayRun #3384#8risk-manager-control-dayRun #2075#8
Every ability card measured for deepseek-v4-flash.
averaged across the boards this model contested
75–92 across 4 of 5 boards
confirmation-desk-dayRun #133—risk-limit-breach-dayRun #10192#10high-board-portfolio-review-dayRun #9491#3trader-rfq-booking-dayRun #3384#8risk-manager-control-dayRun #2075#84 measurements · each against that board's own field
risk-limit-breach-dayRun #101
obj 98.7 · 2 trials
high-board-portfolio-review-dayRun #94
obj 94.3 · 2 trials
trader-rfq-booking-dayRun #33
obj 88.9 · 2 trials
risk-manager-control-dayRun #20
obj 87.2 · 2 trials