Acropolis

Qwen2.5 VL 32B Instruct

AlibabaGrade B. Evidence grade B of A to D: good.Open weightsConsistent across benchmarks
Compare with peers

Domain profile

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

  • Agentic (1 boards): 0 (scale 0 to 100)Agentic (1 boards)0
  • Vision (3 boards): 22 (scale 0 to 100)Vision (3 boards)22

Benchmark appearances

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

BenchmarkRankScoreCostStatusObserved
OSWorld

Epoch AI Benchmarking Hub · Default

#37 / 38
3.9%UnknownStandardNot stated
OSWorld

Epoch AI Benchmarking Hub · Default

#38 / 38
3%UnknownStandardNot stated
LMArena Vision

LMArena · English · Default

#119 / 167
1,158[1,139 to 1,178]UnknownStandard9 Nov 2026
LMArena Vision

LMArena · Overall · Default

#119 / 167
1,152[1,137 to 1,167]UnknownStandard9 Nov 2026
LMArena Vision (style control)

LMArena · English · Default

#133 / 167
1,123[1,103 to 1,142]UnknownStandard9 Nov 2026
LMArena Vision (style control)

LMArena · Overall · Default

#133 / 167
1,118[1,103 to 1,134]UnknownStandard9 Nov 2026

Closest head-to-head rivals

Direct pairwise win rates for Qwen2.5 VL 32B Instruct across shared benchmark boards.