GPT-5.2 vs Claude Opus 4 (2025), and what replaced them

This page first compared OpenAI’s GPT-5.2 with Anthropic’s Claude Opus 4. Both models have since been replaced: on LazySusan, GPT-5.2 conversations now run on GPT-6.1 Sol and Claude Opus 4.x on Claude Opus 5.5. The specific scores and limits we once listed no longer describe anything you can use, so we have removed them. What follows is the part of the comparison that has held up across generations.

Summary of where each option has the edge
Area	Edge
Writing and long documents	Claude
Everyday tasks and speed	GPT
Code	Tie
Where to read the current comparison	—

Area

Edge

Writing and long documents

Claude

Everyday tasks and speed

GPT

Code

Tie

Where to read the current comparison

—

Writing and long documents

Claude’s strength in 2025 was careful, natural writing and reading long material closely, and that is still how the current Claude models are best used: long documents, careful analysis and code review. OpenAI’s models were, and remain, quick and versatile for everyday writing.

edge: Claude

Everyday tasks and speed

For short, varied tasks such as quick answers, brainstorming and reformatting, OpenAI’s default model was the faster all-rounder in 2025. GPT-6.1 Sol plays the same role today, with GPT-6 Luna for the cheapest quick replies.

edge: GPT

Code

Both families handle code well. The pattern we saw then still applies: Claude tends to do best when it can read many files at once and trace a problem across them, while OpenAI’s models are quick for focused fixes and snippets.

edge: Tie

Where to read the current comparison

Model families now move several versions a year, so we keep head-to-head pages dated. The current OpenAI versus Anthropic comparison covers GPT-6 Astra, GPT-6.1 Sol, Claude Opus 5.5 and Claude Fable 5.1, with what each costs on LazySusan.

Writing and long documents

Everyday tasks and speed

Code

Where to read the current comparison

Every leading model, one workspace.