Claude Opus 4.8vsGLM-5.3

Claude Opus 4.8
GLM-5.3
Specifications
Parameters
743B
Context window
1M
Benchmarks
Nonsense detection
BullshitBench v2
95%
Prompt injection robustness
Gray Swan IPI · k = 1
0.5%
Prompt injection robustness
Gray Swan IPI · k = 10
4.1%
Prompt injection robustness
Gray Swan IPI · k = 15
5.5%
Agentic coding
SWE-Bench Pro
69.2%
Multilingual coding
SWE-Bench Multilingual
84.4%
Agentic coding
CursorBench v3.2
62.3%
Agentic coding
CursorBench v3.1
63.8%
Agentic coding
DeepSWE 1.1
66.9%
Agentic coding
DeepSWE 1.0
55.8%
Next.js coding
Next.js Evals
88%
Agentic computer work
Frontier-Bench v0.1
21.1%
Agentic terminal coding
Terminal-Bench 3.0
28.3%
Agentic terminal coding
Terminal-Bench 2.1
74.6%
Browser agent
BU Bench
74%
Web browsing
BrowseComp
84.3%
Cybersecurity
CyberGym
84.5%
Cybersecurity
ExploitBench
54.4%
Cybersecurity
ExploitGym · 6-hour budget
130
Cybersecurity
ExploitGym · 2-hour budget
105
Multidisciplinary reasoning
Humanity's Last Exam · no tools
49.8%
Multidisciplinary reasoning
Humanity's Last Exam · with tools
57.9%
62.5%Best
Agentic computer use
OSWorld-Verified
83.4%
Agentic computer use
Agent's Last Exam
28.5%
Business workflows
AutomationBench
48.2%
Agentic financial analysis
Finance Agent v2
53.9%
Agentic legal work
Harvey's Legal Agent Benchmark
9.58%
Tax questions
TaxEval v2
75.63%
Medical admin work
MedScribe
85.75%
Knowledge work
GDPval-AA
1890
Knowledge work
GDPval-AA v2
1600
1769Best
Community preference
Arena Elo (Text)
1482
Community preference (code)
Arena Elo (Code)
1568
Overview
CompanyAnthropicZ.ai
Release dateMay 28 2026Aug 14 2026
AccessProprietaryProprietary

Which is better: Claude Opus 4.8 or GLM-5.3?

GLM-5.3 leads Claude Opus 4.8 on 2 of the 2 benchmarks they both report (Humanity's Last Exam, GDPval-AA v2). Claude Opus 4.8 shipped 78 days before GLM-5.3, so benchmark comparisons should account for the intervening progress.

Published specifications for these two models are limited — see each model page for the latest details.

On Humanity's Last Exam · with tools, GLM-5.3 leads at 62.5% vs Claude Opus 4.8 at 57.9%. On GDPval-AA v2, GLM-5.3 leads at 1769 vs Claude Opus 4.8 at 1600.

Frequently asked questions

Claude Opus 4.8 was released by Anthropic on May 28 2026.

GLM-5.3 was released by Z.ai on Aug 14 2026.

GLM-5.3 leads on Humanity's Last Exam · with tools — Claude Opus 4.8 57.9% vs GLM-5.3 62.5%.

Other comparisons