Noometry

Model comparison

GLM-5.3 vs Grok 4

GLM-5.3 is the stronger model overall, scoring 54.8 to 48.1 on the Noometry Index.

Last verified . 22 shared benchmarks.

GLM-5.3 Z.ai (Zhipu)

54.8

Rank #26 Confirmed

Grok 4 xAI

48.1

Rank #56 Confirmed

Summary

  • They share 22 benchmarks with published results for both. GLM-5.3 scores higher in 7 categories and Grok 4 in 2 categories; 9 gaps are clear of the uncertainty.
  • The widest gap is in long context, where Grok 4 leads 63.1 to 45.4.
  • The biggest single-benchmark swing is WeirdML: 75.4% for GLM-5.3 and 45.7% for Grok 4.
  • GLM-5.3 has downloadable open weights; the other is API-only.

Side by side

GLM-5.3Grok 4
ProviderZ.ai (Zhipu)xAI
Noometry Index54.848.1
Released2026-08-142025-07-09
WeightsOpenProprietary
Context window1M—
Max output131K—
Input $ / M tokens$1.40—
Output $ / M tokens$4.40—
Results tracked4248

Sponsored placements are available on pages like this one. Advertise on Noometry

Category by category

Coding GLM-5.3 leads

GLM-5.3: 59.5 (#14), Grok 4: 50.3 (#46)

BenchmarkGLM-5.3Grok 4
WeirdML75.4%45.7%
LMArena Coding14961408
DeepSWE69%—
FrontierCode40.1%—
Aider Polyglot—79.6%
CursorBench42.6%—
LMArena WebDev1622—
FrontierSWE30.2%—
SciCode59%—
ALE-Bench1,317—

Agentic & Tool Use GLM-5.3 leads

GLM-5.3: 36.4 (#38), Grok 4: 32.3 (#68)

BenchmarkGLM-5.3Grok 4
Terminal-Bench—27.2%
APEX-Agents56.6%—
Berkeley Function Calling Leaderboard—63%
GDPval—21.1%
Cybench—43%
DeepResearch Bench—47.3%
BALROG—43.6%
LMArena Search—1142
METR Time Horizons—66.6%
Vending-Bench 28,164—

Reasoning GLM-5.3 leads

GLM-5.3: 46.1 (#46), Grok 4: 36.7 (#65)

BenchmarkGLM-5.3Grok 4
Chess Puzzles21%28%
LMArena Hard Prompts14891409
Epoch Capabilities Index155.61146.44
ARC-AGI-2—16%
SimpleBench—60.5%
Kagi LLM Benchmark—73.6%
NYT Connections (extended)74.2%—
ARC-AGI-1—66.7%
CritPt19.1%—
Mystery Game Puzzles33%—
DTBench87.7%—
LMCA55.5%—
Bench to the Future 30.15—
ForecastBench—60.9

Math GLM-5.3 leads

GLM-5.3: 62.3 (#33), Grok 4: 48.4 (#64)

BenchmarkGLM-5.3Grok 4
OTIS Mock AIME 2024-202591.1%84%
LMArena Math14891422
FrontierMath (Tiers 1-3)68.8%—
FrontierMath Tier 429.3%—
ProofBench49%—
Omni-MATH—60.3%
FrontierMath (Feb 2025 set)—19.7%
FrontierMath Tier 4 (v1)—2.1%

Knowledge GLM-5.3 leads

GLM-5.3: 58.3 (#37), Grok 4: 53.8 (#55)

BenchmarkGLM-5.3Grok 4
GPQA Diamond90.9%87%
LMArena Expert15161415
SimpleQA Verified41%—
MMLU-Pro—85.1%
Confabulations—12.4%
GPQA (HELM)—72.7%

Multimodal Not comparable

GLM-5.3: —, Grok 4: 33.7 (#94)

BenchmarkGLM-5.3Grok 4
LMArena Vision—1210
GeoBench—45%

Multilingual GLM-5.3 leads

GLM-5.3: 55.7 (#28), Grok 4: 51.8 (#103)

BenchmarkGLM-5.3Grok 4
LMArena Non-English14571403
LMArena Chinese15281427
LMArena French14991418
LMArena German14991429
LMArena Japanese14531394
LMArena Korean14721377
LMArena Russian14631410
LMArena Spanish14601420

Instruction Following Grok 4 leads

GLM-5.3: 77.5 (#23), Grok 4: 79.2 (#5)

BenchmarkGLM-5.3Grok 4
LMArena Instruction Following14771387
IFEval—94.9%

Long Context Grok 4 leads

GLM-5.3: 45.4 (#41), Grok 4: 63.1 (#4)

BenchmarkGLM-5.3Grok 4
LMArena Longer Query14821409
Fiction.LiveBench—94.4%

Writing & Preference GLM-5.3 leads

GLM-5.3: 75.7 (#6), Grok 4: 58.5 (#116)

BenchmarkGLM-5.3Grok 4
LMArena Text14711411
LMArena Creative Writing14571397
LMArena Multi-Turn14721416
Short-Story Creative Writing—76.9%
EQ-Bench Creative Writing2075—
WildBench—79.7%

Frequently asked questions

Is GLM-5.3 better than Grok 4?

GLM-5.3 is the stronger model overall, scoring 54.8 to 48.1 on the Noometry Index.

Is GLM-5.3 or Grok 4 better for coding?

GLM-5.3 scores higher on coding benchmarks: 59.5 versus 50.3 in the Noometry coding category.

How many benchmarks do GLM-5.3 and Grok 4 share?

22 benchmarks have published results for both models. GLM-5.3 has 42 scored results on Noometry and Grok 4 has 48.

Related comparisons

Go deeper