Acropolis

Hy4 Preview

TencentGrade A. Evidence grade A of A to D: strongest.Weights unstatedConsistent across benchmarks
Compare with peers

Domain profile

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

  • Agentic (11 boards): 32 (scale 0 to 100)Agentic (11 boards)32
  • Coding (7 boards): 27 (scale 0 to 100)Coding (7 boards)27
  • Finance (4 boards): 42 (scale 0 to 100)Finance (4 boards)42
  • Legal (3 boards): 32 (scale 0 to 100)Legal (3 boards)32

Benchmark appearances

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

BenchmarkRankScoreCostStatusObserved
Legal Research

Vals AI benchmarks · Default

#16 / 38
45.2%[38.4% to 52%]$0.85Standard15 Nov 2026
Tax Agent

Vals AI benchmarks · Default

#17 / 38
32.6%[26% to 39.2%]$0.59Standard15 Nov 2026
Finance Agent

Vals AI benchmarks · Default

#21 / 38
61.6%[57.6% to 65.7%]$0.67Standard15 Nov 2026
EMB

Vals AI benchmarks · Default

#24 / 38
68.9%[64.2% to 73.7%]$0.79Standard15 Nov 2026
LMArena WebDev

LMArena · Overall · Default

#26 / 156
1,632[1,622 to 1,641]UnknownStandard15 Nov 2026
LMArena WebDev

LMArena · Webdev · Default

#26 / 156
1,632[1,622 to 1,641]UnknownStandard15 Nov 2026
LMArena WebDev

LMArena · Webdev react · Default

#26 / 140
1,653[1,640 to 1,665]UnknownStandard15 Nov 2026
Legal Agent Benchmark (HLAB-Vals)

Vals AI benchmarks · Default

#26 / 38
8.8%[4.5% to 13%]$0.98Standard15 Nov 2026
LMArena WebDev

LMArena · Webdev HTML · Default

#27 / 155
1,624[1,599 to 1,650]UnknownStandard15 Nov 2026
Code Migration

Vals AI benchmarks · Default

#28 / 38
44.7%[36.1% to 53.3%]$3.42Standard15 Nov 2026
Vals Index

Vals AI benchmarks · Default

#28 / 38
53.2%[51% to 55.4%]$1.57Standard15 Nov 2026
Vibe Code Bench

Vals AI benchmarks · OpenHands

#28 / 38
77.5%[69.6% to 85.4%]$2.18Standard15 Nov 2026
Terminal-Bench 4

Vals AI benchmarks · Default

#32 / 38
8.1%[5.5% to 10.7%]$3.07Standard15 Nov 2026
LMArena Agent

LMArena · Overall · Default

#34 / 64
-0[-0 to 0]$0.43Standard15 Nov 2026
Bug Hunt Bench

Bug Hunt Bench · Claude Code

#97 / 145
18$3.13Standard28 Aug 2026

Closest head-to-head rivals

Direct pairwise win rates for Hy4 Preview across shared benchmark boards.