Acropolis

Mercury 3

InceptionGrade C. Evidence grade C of A to D: thin.Weights unstatedConsistent across benchmarks
Compare with peers

Domain profile

Median within-board percentile by capability domain across boards with sufficient peer overlap.

  • Agentic (9 boards): 11 (scale 0 to 100)Agentic (9 boards)11
  • Coding (4 boards): 11 (scale 0 to 100)Coding (4 boards)11
  • Finance (4 boards): 11 (scale 0 to 100)Finance (4 boards)11
  • Legal (3 boards): 12 (scale 0 to 100)Legal (3 boards)12

Benchmark appearances

Best verified run per board, with confidence intervals and Pareto frontier status.

BenchmarkRankScoreCostStatusObserved
Legal Agent Benchmark (HLAB-Vals)

Vals AI benchmarks · high

#30 / 38
5%[1.8% to 8.2%]$0.19Standard15 Nov 2026
Code Migration

Vals AI benchmarks · high

#32 / 38
36.9%[29.7% to 44%]$2.21Standard15 Nov 2026
Legal Research

Vals AI benchmarks · high

#33 / 38
28.8%[22.7% to 35%]$0.07Standard15 Nov 2026
Tax Agent

Vals AI benchmarks · high

#33 / 38
18.2%[13.2% to 23.1%]$0.05Standard15 Nov 2026
Finance Agent

Vals AI benchmarks · high

#34 / 38
54.7%[50.8% to 58.6%]$0.06Standard15 Nov 2026
Vals Index

Vals AI benchmarks · high

#34 / 38
43.7%[41.3% to 46.1%]$0.48Standard15 Nov 2026
Vibe Code Bench

Vals AI benchmarks · high · OpenHands

#34 / 38
60.9%[50.4% to 71.5%]$0.63Standard15 Nov 2026
EMB

Vals AI benchmarks · high

#35 / 38
57.9%[52.2% to 63.6%]$0.19Standard15 Nov 2026
Terminal-Bench 4

Vals AI benchmarks · high

#35 / 38
3%[3% to 3%]$0.42Standard15 Nov 2026

Closest head-to-head rivals

Direct pairwise win rates for Mercury 3 across shared benchmark boards.