GLM-5.3 Benchmark Details

GLM-5.3 currently shows benchmark results led by HLE (3 / 176, score 62.50), Terminal-Bench 2.1 (3 / 38, score 88.20), Automation Bench (1 / 4, score 48.20).

Benchmark Results

GLM-5.3

Benchmark Results

General Knowledge

1 evaluations

Benchmark / mode

Score

Rank/total

HLE

MaxTools

62.50

3 / 176

AI Agent - Tool Usage

8 evaluations

Benchmark / mode

Score

Rank/total

ExploitGym (6h)

MaxTools

130

1 / 1

ExploitGym (2h)

MaxTools

105

1 / 1

Terminal-Bench 2.1

MaxTools

88.20

3 / 38

CyberGym

MaxTools

84.50

1 / 2

Toolathlon-Verified

MaxTools

2 / 3

ExploitBench

MaxTools

54.40

1 / 1

Automation Bench

MaxTools

48.20

1 / 4

Terminal-Bench 3.0

MaxTools

28.30

1 / 2

Coding and Software Engineer

6 evaluations

Benchmark / mode

Score

Rank/total

FrontierSWE

MaxTools

78.10

2 / 2

DeepSWE

MaxTools

66.90

8 / 22

NL2Repo-Bench

MaxTools

1 / 2

SWE-Marathon

MaxTools

42.50

1 / 3

PostTrain Bench

MaxTools

39.80

1 / 2

Program Bench

MaxTools

2 / 2

Agent Level Benchmark

1 evaluations

Benchmark / mode

Score

Rank/total

Agents' Last Exam

MaxTools

28.50

4 / 6

Productivity Knowledge

1 evaluations

Benchmark / mode

Score

Rank/total

GDPval-AA v2

MaxTools

1769

2 / 8

Compare with other models