Acropolis

Hy4

TencentGrade B. Evidence grade B of A to D: good.Weights unstatedReleased 3 Nov 2026Consistent across benchmarks
Compare with peers

Domain profile

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

  • Agentic (10 boards): 60 (scale 0 to 100)Agentic (10 boards)60
  • Coding (5 boards): 64 (scale 0 to 100)Coding (5 boards)64
  • Finance (4 boards): 60 (scale 0 to 100)Finance (4 boards)60
  • Legal (3 boards): 54 (scale 0 to 100)Legal (3 boards)54

Benchmark appearances

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

BenchmarkRankScoreCostStatusObserved
Code Migration

Vals AI benchmarks · Default

#14 / 38
59.1%[52.6% to 65.6%]$3.24Standard15 Nov 2026
Finance Agent

Vals AI benchmarks · Default

#14 / 38Tied #1
64.7%[60.7% to 68.7%]$1.28Standard15 Nov 2026
Terminal-Bench 4

Vals AI benchmarks · Default

#14 / 38
39.4%[37.7% to 41.1%]$1.28Standard15 Nov 2026
Tax Agent

Vals AI benchmarks · Default

#15 / 38
34.4%[27.9% to 40.9%]$1.69Standard15 Nov 2026
Vals Index

Vals AI benchmarks · Default

#17 / 38
59.4%[57.2% to 61.5%]$1.72Standard15 Nov 2026
LMArena Agent

LMArena · Overall · Default

#18 / 64
0.1[0 to 0.1]$0.57Standard15 Nov 2026
Legal Agent Benchmark (HLAB-Vals)

Vals AI benchmarks · Default

#18 / 38
12.5%[7.3% to 17.7%]$1.04Standard15 Nov 2026
Legal Research

Vals AI benchmarks · Default

#20 / 38
41.3%[34.6% to 48.1%]$2.22Standard15 Nov 2026
Vibe Code Bench

Vals AI benchmarks · OpenHands

#24 / 38
81%[74.1% to 88%]$2.39Standard15 Nov 2026
EMB

Vals AI benchmarks · Default

#30 / 38
66.2%[61% to 71.4%]$0.62Standard15 Nov 2026

Closest head-to-head rivals

Direct pairwise win rates for Hy4 across shared benchmark boards.