GLM-5.2 (mock model card)

[!NOTE] This is a mock repository. It exists to showcase how far you can push a Hugging Face model card with sanitized inline-style HTML. The chart below is a pure-HTML recreation of the benchmark figure from the zai-org/GLM-5.2 model card, using Hugging Face org avatars as the chart logos. All numbers are illustrative.

LLM Performance Evaluation

8 Benchmarks: SWE-bench Pro, Terminal-Bench 2.1 (Terminus), NL2Repo, DeepSWE, ProgramBench, MCP-Atlas, Tool-Decathlon, Humanity’s Last Exam

All models are evaluated under their maximum thinking effort

GLM-5.2GLM-5.1Claude Opus 4.8GPT-5.5Gemini 3.1 Pro

62.1

58.4

69.2

58.6

54.2

SWE-bench Pro

81.0

63.5

85.0

84.0

74.0

Terminal-Bench 2.1

48.9

42.7

69.7

50.7

33.4

NL2Repo

46.2

18.0

58.0

70.0

10.0

DeepSWE

63.7

50.9

71.9

70.8

39.5

ProgramBench

77.0

71.8

77.8

75.3

69.2

MCP-Atlas

48.2

40.7

59.9

55.6

48.8

Tool-Decathlon

54.7w/ Tools

40.5

52.3w/ Tools

31.0

57.9w/ Tools

49.8

52.2w/ Tools

41.4

51.4w/ Tools

45.0

Humanity's Last Exam

How this chart is built