Acropolis

Qwen3.8 Max Internal

AlibabaGrade 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): 60 (scale 0 to 100)Agentic (9 boards)60
  • Coding (4 boards): 56 (scale 0 to 100)Coding (4 boards)56
  • Finance (4 boards): 61 (scale 0 to 100)Finance (4 boards)61
  • Legal (3 boards): 65 (scale 0 to 100)Legal (3 boards)65

Benchmark appearances

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

BenchmarkRankScoreCostStatusObserved
Tax Agent

Vals AI benchmarks · Default

#3 / 38Tied #1
49.7%[42.6% to 56.7%]$4.07Standard15 Nov 2026
Legal Agent Benchmark (HLAB-Vals)

Vals AI benchmarks · Default

#11 / 38
15.4%[9.4% to 21.4%]$3.36Standard15 Nov 2026
EMB

Vals AI benchmarks · Default

#13 / 38Tied #1
72%[67.3% to 76.7%]$4.43Standard15 Nov 2026
Legal Research

Vals AI benchmarks · Default

#13 / 38
46.2%[39.4% to 52.9%]$5.31Standard15 Nov 2026
Code Migration

Vals AI benchmarks · Default

#16 / 38
54.5%[47.4% to 61.7%]$25.86Standard15 Nov 2026
Terminal-Bench 4

Vals AI benchmarks · Default

#16 / 38
38.9%[36.3% to 41.5%]$12.15Standard15 Nov 2026
Vals Index

Vals AI benchmarks · Default

#18 / 38
59%[56.9% to 61.2%]$8.73Standard15 Nov 2026
Finance Agent

Vals AI benchmarks · Default

#29 / 38
58.8%[54.7% to 63%]$2.99Standard15 Nov 2026
Vibe Code Bench

Vals AI benchmarks · OpenHands

#29 / 38
77.1%[69.8% to 84.4%]$11.67Standard15 Nov 2026

Closest head-to-head rivals

Direct pairwise win rates for Qwen3.8 Max Internal across shared benchmark boards.