GPT-5.6 Sol vs Claude Fable 5 — Who Actually Wins? Real Benchmark Data (2026)

head to head

GPT-5.6 Sol vs Claude Fable 5

30 shared benchmarks

11

wins

0

ties

19

wins

39.9%

APEX-Agents

43.3%

52.7%

Agents' Last Exam

40.5%

29.7%

AutomationBench

29.1%

88.9%

BabyVision

90.5%

90.4%

BrowseComp

88.0%

84.6%

CharXiv

88.9%

72.7%

DeepSWE

70.0%

73.5%

ExploitBench

78.0%

83.0%

FrontierMath

87.0%

71.3%

FrontierSWE

86.6%

94.6%

GPQA Diamond

92.6%

44.5%

Humanity's Last Exam

53.3%

46.5%

JobBench

57.4%

95.8%

MATH-Vision

94.8%

83.6%

MCP-Atlas

84.7%

46.2%

MLS Bench

49.9%

83.0%

MMMU Pro

81.2%

63.2%

OfficeQA Pro

69.9%

85.8%

OmniDocBench

89.8%

59.7%

PerceptionBench

57.2%

34.6%

PostTrain Bench

41.4%

77.6%

Program Bench

76.8%

39.0%

SWE Marathon

35.0%

64.6%

SWE-bench Pro

80.3%

32.4%

SpreadsheetBench 2

34.7%

88.8%

Terminal-Bench 2.1

84.6%

58.0%

Toolathlon

61.7%

74.9%

Toolathlon-Verified

77.9%

41.8%

WorldVQA

56.7%

17.0%

ZeroBench

23.0%