ChatGPT or Claude - which is better for writing and code?

On the available evidence, Claude looks slightly ahead for both writing and code, but the margins are small and much of the coding data comes from Anthropic itself. ChatGPT Plus and Claude Pro both cost 20 USD per month, so price does not settle it.

• Writing: on the LMArena text leaderboard of 27 October 2026, Claude Opus 5.5 ranks third overall, and no GPT-6 model is in the top five.
• Code benchmarks: Anthropic reports Claude Opus 5.5 at 66.4 percent on Terminal-Bench 4.0 versus 57.9 percent for GPT-6 Astra, and 54.4 percent versus 53.3 percent on FrontierCode v1.1.
• Read those with care: these are vendor-published numbers. Artificial Analysis lists 59.6 percent on Terminal-Bench for Opus 5.5, so only compare figures from the same source.
• Coding agents: Claude Code runs in the terminal, VS Code, JetBrains IDEs, a desktop app and on the web, but it is not included in the free plan. OpenAI's Codex comes with ChatGPT Plus, Pro and Business plans.
• API costs: Claude Sonnet 5.5 and GPT-6.1 Sol both cost 2 / 10 USD per 1 million tokens (input / output).

Both services have a free tier, so the most reliable test is to give each one the same real task from your own work, such as a text you need to edit or a bug you need to fix, and compare the results side by side.