Lech Mazur Writing ↗7.34 / 10
GLM-4.5 · Lech Mazur Writing · Mean rubric score / 10 ↗
Z.ai / MODEL EVIDENCE
GLM-4.5
16 benchmarks · 16 observations · 16 sources
Scores at a glance
16 benchmarks · large dot: GLM-4.5; small dots: closest models where reported
WeirdML (v2) ↗40.58%
GLM-4.5 Thinking · WeirdML · Accuracy · % ↗
GLM-4.5 · HLE · Accuracy · % ↗
LegalBench ↗75.63 %
GLM 4.5 · LegalBench · Accuracy ↗
CorpFin v2 ↗60.96 %
GLM 4.5 · CorpFin · Accuracy ↗
GLM 4.5 · TaxEval · Accuracy ↗
MedQA (USMLE) ↗89.97 %
GLM 4.5 · MedQA · Accuracy ↗
AIME 2024–2025 · Vals ↗86.67 %
GLM 4.5 · AIME · Accuracy ↗
MATH-500 ↗94 %
GLM 4.5 · MATH 500 · Accuracy ↗
MGSM (Multilingual GSM) ↗90.84 %
GLM 4.5 · MGSM · Accuracy ↗
GPQA Diamond ↗72.22 %
GLM 4.5 · GPQA Diamond · Accuracy ↗
MMLU-Pro ↗81.22 %
GLM 4.5 · MMLU Pro · Accuracy ↗
LiveCodeBench ↗67.45 %
GLM 4.5 · LiveCodeBench · Accuracy ↗
glm-4.5 · FrontierMath-2025-02-28-Public · FrontierMath public · 10 questions · source run score · Epoch AI · public February 2025 sample · historical release ↗
AlgoTune ↗1.52 ×
GLM-4.5 (high) · Harmonic-mean speedup ↗
ALE-Bench ↗344.82 Performance points
glm-4.5 · Average contest performance ↗
Closest models by score
Similar positions on shared benchmark leaderboards. Smaller gaps are closer.
9 of 10 shared benchmark comparisons
Qwen 3 (235B) →Qwen
9 of 10 shared benchmark comparisons
MiMo V2.5 →Xiaomi
6 of 10 shared benchmark comparisons
How similarity is measured
Mean absolute gap in percentile positions on shared benchmark settings; lower is closer. At least 60% overlap is required, with a minimum of three when available. Independent evaluations are used when there are at least three; otherwise published report comparisons are used. Repeated model observations within one setting are averaged before ranking. Writing-style traits and composite indices are excluded. This describes similarity, not an overall quality ranking.
Sources
Every plotted observation links to its evaluation. Reported configurations, benchmark versions and task subsets can differ; similarity describes shared scores, not training lift.
- Ale Bench — methodology and published results ↗
- Algotune — methodology and published results ↗
- Humanity's Last Exam ↗
- Writing benchmark — incorporating mandatory story elements ↗
- WeirdML ↗
- A Rosetta Stone for AI Benchmarks — released Epoch evaluation runs ↗
- AIME — evaluation methodology and results ↗
- CorpFin — evaluation methodology and results ↗
- GPQA Diamond — evaluation methodology and results ↗
- LiveCodeBench — evaluation methodology and results ↗
- LegalBench — evaluation methodology and results ↗
- MATH 500 — evaluation methodology and results ↗
- MedQA — evaluation methodology and results ↗
- MGSM — evaluation methodology and results ↗
- MMLU Pro — evaluation methodology and results ↗
- TaxEval — evaluation methodology and results ↗
Release dates reported by sources
2025-08-03 · 2025-08-05
Names used in source tables
GLM 4.5 · GLM-4.5 · GLM-4.5 (high) · GLM-4.5 Thinking · glm-4.5