Jev-1.13 is a fast model with strong common-sense judgment for System One tasks, but has documented failure modes including literal reading, unreliable counting and math, poor numeric precision, weak date/time reasoning, and difficulty with indirect instructions.
jev-1.13 is a fast model good at common-sense tasks but has documented failure modes including literal interpretation, poor mathematical reasoning, unreliable counting, weak numeric calibration, and difficulty with date comparisons and indirect instructions.