Claude Fable 5 vs GLM-5.2 — Who Actually Wins? Real Benchmark Data (2026)

head to head

Claude Fable 5 vs GLM-5.2

18 shared benchmarks

18

wins

0

ties

0

wins

43.3%

APEX-Agents

35.6%

29.1%

AutomationBench

12.9%

70.0%

DeepSWE

44.0%

86.6%

FrontierSWE

67.3%

92.6%

GPQA Diamond

91.2%

53.3%

Humanity's Last Exam

40.5%

57.4%

JobBench

43.4%

84.7%

MCP-Atlas

76.8%

49.9%

MLS Bench

40.4%

69.9%

OfficeQA Pro

41.4%

41.4%

PostTrain Bench

34.3%

76.8%

Program Bench

63.7%

35.0%

SWE Marathon

13.0%

80.3%

SWE-bench Pro

62.1%

34.7%

SpreadsheetBench 2

28.12%

84.6%

Terminal-Bench 2.1

81.0%

61.7%

Toolathlon

48.2%

77.9%

Toolathlon-Verified

59.9%