Shared benchmarks

25
Shared benchmarks for GLM-5.2 and GLM-5.3
BenchmarkGLM-5.2GLM-5.3
  • 45.2%
  • 56.6%
Evaluation details for APEX-Agents Original

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

GLM-5.2

Record for GLM-5.2 on APEX-Agents Original

GLM-5.3

Record for GLM-5.3 on APEX-Agents Original
AutomationBench 1.0.6Potentially non-equivalent: Evaluator sets differ.
  • 26.2%1.0.6 · Zapier AutomationBench Team
  • 48.2%1.0.6 · Z.ai
Potentially non-equivalent · Evaluation details for AutomationBench 1.0.6

Evaluator sets differ. Evaluation methodology is not recorded by the registry.

GLM-5.2

Record for GLM-5.2 on AutomationBench 1.0.6

GLM-5.3

Record for GLM-5.3 on AutomationBench 1.0.6
  • 77.2%
  • 84.5%
Evaluation details for CyberGym Original

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

GLM-5.2

Record for GLM-5.2 on CyberGym Original

GLM-5.3

Record for GLM-5.3 on CyberGym Original
  • 44.0%
  • 69.0%
Evaluation details for DeepSWE 1.1

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

GLM-5.2

Record for GLM-5.2 on DeepSWE 1.1

GLM-5.3

Record for GLM-5.3 on DeepSWE 1.1
  • 38.9%
  • 70.5%
Evaluation details for DeepSWE 1.1 — Mercor mini-swe-agent, 500 steps

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

GLM-5.2

Record for GLM-5.2 on DeepSWE 1.1 — Mercor mini-swe-agent, 500 steps

GLM-5.3

Record for GLM-5.3 on DeepSWE 1.1 — Mercor mini-swe-agent, 500 steps
  • 49.7%
  • 55.8%
Evaluation details for Finance Agent Benchmark 2 — October 2026 evaluation

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

  • 71.2%none
  • 87.9%low
  • 91.9%max
  • 90.9%
Evaluation details for GPQA Diamond — Epoch AI harness

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

GLM-5.2

Record for GLM-5.2 on Graduate-Level Google-Proof Q&A Diamond — Epoch AI harness
Record for GLM-5.2 on Graduate-Level Google-Proof Q&A Diamond — Epoch AI harness
Record for GLM-5.2 on Graduate-Level Google-Proof Q&A Diamond — Epoch AI harness

GLM-5.3

Record for GLM-5.3 on Graduate-Level Google-Proof Q&A Diamond — Epoch AI harness
  • 90.2%
  • 91.3%
Evaluation details for GPQA Diamond — Mercor single-shot

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

GLM-5.2

Record for GLM-5.2 on Graduate-Level Google-Proof Q&A Diamond — Mercor single-shot

GLM-5.3

Record for GLM-5.3 on Graduate-Level Google-Proof Q&A Diamond — Mercor single-shot
  • 85.6%
  • 88.1%
Evaluation details for GPQA Diamond — Vals AI harness

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

GLM-5.2

Record for GLM-5.2 on Graduate-Level Google-Proof Q&A Diamond — Vals AI harness

GLM-5.3

Record for GLM-5.3 on Graduate-Level Google-Proof Q&A Diamond — Vals AI harness
  • 7.1%
  • 8.3%
Evaluation details for HLAB October 2026 evaluation

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

GLM-5.2

Record for GLM-5.2 on Harvey's Legal Agent Benchmark October 2026 evaluation

GLM-5.3

Record for GLM-5.3 on Harvey's Legal Agent Benchmark October 2026 evaluation
  • 31.7%
  • 40.0%
Evaluation details for Humanity's Last Exam Text only — Mercor single-shot

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

GLM-5.2

GLM-5.3

  • 77.8%
  • 84.2%
Evaluation details for MCP Atlas Public, April 2026 update

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

GLM-5.2

Record for GLM-5.2 on MCP Atlas Public, April 2026 update

GLM-5.3

Record for GLM-5.3 on MCP Atlas Public, April 2026 update
  • 85.9%
  • 88.0%
Evaluation details for MMLU-Pro Original — Mercor single-shot

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

GLM-5.2

Record for GLM-5.2 on MMLU-Pro Original — Mercor single-shot

GLM-5.3

Record for GLM-5.3 on MMLU-Pro Original — Mercor single-shot
  • 86.7%
  • 86.8%
Evaluation details for MMLU-Pro Original — Vals AI harness

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

GLM-5.2

Record for GLM-5.2 on MMLU-Pro Original — Vals AI harness

GLM-5.3

Record for GLM-5.3 on MMLU-Pro Original — Vals AI harness
  • 48.9%
  • 58.0%
Evaluation details for NL2Repo Bench Original

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

GLM-5.2

Record for GLM-5.2 on NL2Repo Bench Original

GLM-5.3

Record for GLM-5.3 on NL2Repo Bench Original
  • 30.4%
  • 34.5%
Evaluation details for PostTrainBench 1.2 — Claude Code

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

GLM-5.2

Record for GLM-5.2 on PostTrainBench 1.2 — Claude Code

GLM-5.3

Record for GLM-5.3 on PostTrainBench 1.2 — Claude Code
  • 10.4%
  • 12.0%
Evaluation details for Riemann-bench October 2026 leaderboard

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

GLM-5.2

Record for GLM-5.2 on Riemann-bench October 2026 leaderboard

GLM-5.3

Record for GLM-5.3 on Riemann-bench October 2026 leaderboard
  • 34.2%
  • 41.0%
Evaluation details for SimpleQA Verified Original — Epoch AI harness

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

  • 46.0%
  • 42.5%
Evaluation details for SWE Atlas Codebase QnA — Mercor mini-swe-agent, 250 steps

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

  • 78.6%
  • 87.6%
Evaluation details for SWE-bench Multilingual — Mercor mini-swe-agent, 1000 steps

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

  • 78.5%
  • 81.7%
Evaluation details for SWE-bench Verified — Mercor mini-swe-agent, 1000 steps

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

GLM-5.2

GLM-5.3

  • 82.8%
  • 95.4%
Evaluation details for SWE-bench Verified — Vals AI mini-swe-agent

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

GLM-5.2

Record for GLM-5.2 on SWE-bench Verified — Vals AI mini-swe-agent

GLM-5.3

Record for GLM-5.3 on SWE-bench Verified — Vals AI mini-swe-agent
  • 67.8%
  • 71.5%
Evaluation details for Terminal-Bench 2.1 — Vals AI Terminus 2

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

GLM-5.2

Record for GLM-5.2 on Terminal-Bench 2.1 — Vals AI Terminus 2

GLM-5.3

Record for GLM-5.3 on Terminal-Bench 2.1 — Vals AI Terminus 2
Toolathlon VerifiedPotentially non-equivalent: Evaluator sets differ.
  • 59.9%Verified · Toolathlon Authors
  • 73.0%Verified · Z.ai
Potentially non-equivalent · Evaluation details for Toolathlon Verified

Evaluator sets differ. Evaluation methodology is not recorded by the registry.

GLM-5.2

Record for GLM-5.2 on Toolathlon Verified

GLM-5.3

Record for GLM-5.3 on Toolathlon Verified
  • 64.0%
  • 78.1%
Evaluation details for Vibe Code Bench 1.1 — OpenHands, October 2026 evaluation

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.