Beyond the model, what do you actually get for your money?

The model is only half the deal. The other half is everything wrapped around it, and the two camps disagree about what belongs there.

Voice used to be the easy dividing line. Not anymore. Claude's voice mode is in beta on every plan, and ChatGPT simply had a head start, so today it is a maturity gap, not a yes or no.

The rest is philosophy. ChatGPT goes wide: custom GPTs, Deep Research for long multi-step investigations, image generation, and video with Sora on the higher tiers. If you want one subscription that covers text, images, voice and research, it packs in the most.

Claude goes narrow and deep. Artifacts open a live side panel where code, documents and diagrams get built in front of you. Projects keep files, chats and instructions in one workspace. MCP plugs it into outside tools. And Claude Code is a coding agent that lives in your terminal. Reviewers keep praising its writing quality and the way it sticks to detailed instructions.

But hold on, coding agents come bundled everywhere now, right? So does any of this even matter?

It matters less than it used to. Both sides now ship a coding agent with the cheaper plans, Claude Code on Pro and Codex even on the free tier.

Which is why the choice increasingly comes down to how you work, not a feature checklist. Breadth in one place, or focus on writing and careful code.