Acropolis

Grok 4.20

xAIGrade C. Evidence grade C of A to D: thin.ProprietaryReleased 17 Feb 2026Consistent across benchmarks
Compare with peers

Domain profile

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

  • Agentic (2 boards): 57 (scale 0 to 100)Agentic (2 boards)57
  • Coding (1 boards): 68 (scale 0 to 100)Coding (1 boards)68
  • Finance (1 boards): 46 (scale 0 to 100)Finance (1 boards)46
  • General text (1 boards): 69 (scale 0 to 100)General text (1 boards)69
  • Science (1 boards): 69 (scale 0 to 100)Science (1 boards)69

Benchmark appearances

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

BenchmarkRankScoreCostStatusObserved
Vending-Bench 2

Epoch AI Benchmarking Hub · Default

#39 / 72
$4,662.85UnknownStandardNot stated
Terminal Bench

Epoch AI Benchmarking Hub · Grok CLI

#58 / 202
57.3%UnknownStandard10 Mar 2026
Terminal Bench

Epoch AI Benchmarking Hub · grok-cli

#59 / 202
57.3%UnknownStandard2 Apr 2026
ARC-AGI-2

Epoch AI Benchmarking Hub · Default

#90 / 299
65.1%$0.92StandardNot stated

Closest head-to-head rivals

Direct pairwise win rates for Grok 4.20 across shared benchmark boards.