Every few months a release invites the question of whether the ChatGPT/Claude split still means anything. Opus 5, out on 24 July, is a reasonable moment to ask it again — and the answer is that the split held, and arguably got sharper.
The split, stated plainly
ChatGPT is the better default for breadth: it does more things adequately, in more modalities, with the largest ecosystem of integrations and the shortest path from question to answer.
Claude is the better default for depth on text: long documents, careful reasoning, work where tone and nuance carry weight and a confidently wrong answer is expensive.
What Opus 5 changes
Anthropic positioned Opus 5 around long-running agents. That is a depth move, not a breadth move. It reinforces the existing division rather than challenging it.
If you were choosing ChatGPT for its range of modalities and integrations, nothing about this release should move you. If you were choosing Claude because your work is long and expensive to get wrong, you just got more of the reason you chose it.
The uncomfortable truth about picking one
Most teams that can afford both use both, and the split is not usually along the lines the marketing suggests. It is along the lines of which one each person got used to.
If you are standardising on one, the deciding question is rarely capability. It is what the rest of your stack already touches, and what your team will actually open by default at 9am.