Claude Opus 4.8 vs GLM-5.2 — Who Actually Wins? Real Benchmark Data (2026)

head to head

Claude Opus 4.8 vs GLM-5.2

21 shared benchmarks

17

wins

0

ties

4

wins

95.7%

AIME

99.2%

39.4%

APEX-Agents

35.6%

27.2%

AutomationBench

12.9%

59.0%

DeepSWE

44.0%

66.7%

FrontierSWE

67.3%

91.0%

GPQA Diamond

91.2%

96.7%

HMMT Feb 2026

92.5%

96.5%

HMMT Nov 2025

94.4%

49.8%

Humanity's Last Exam

40.5%

48.4%

JobBench

43.4%

77.8%

MCP-Atlas

76.8%

42.8%

MLS Bench

40.4%

63.9%

OfficeQA Pro

41.4%

34.1%

PostTrain Bench

34.3%

71.9%

Program Bench

63.7%

40.0%

SWE Marathon

13.0%

69.2%

SWE-bench Pro

62.1%

31.55%

SpreadsheetBench 2

28.12%

84.6%

Terminal-Bench 2.1

81.0%

59.9%

Toolathlon

48.2%

76.2%

Toolathlon-Verified

59.9%