Claude’s merged chat and Cowork vs. ChatGPT’s Work mode: ChatGPT is faster, Claude is more thorough
The New Stack Jessica Wachtel
Anthropic merged Claude’s chat and Cowork into one app, but a test says ChatGPT Work mode is still quicker. Claude answered with more detail, though, and that extra context may be the real edge.
Based on reporting by The New Stack, Jessica Wachtel — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Anthropic has folded Claude chat and Cowork into a single interface, so users no longer have to guess which tab to click before asking for help. The app now routes the request itself, which is a sensible fix for a problem Anthropic says customers kept running into. The unified setup is going to Pro and Max subscribers first, with free and team users coming later.
OpenAI has been pushing a similar idea since earlier this summer. It launched Work mode on July 9 alongside Chat, then retired the older Agent mode the next month. Work mode lives in the same window as Chat and adds a browser, code execution, and file output, though it still can’t hand a live logged-in browser session back to the user the way Agent mode could for things like logins or payments.
To see how the two systems stack up, the source ran three tests that mirrored real developer work: API research, a small command-line parser, and a handoff from reasoning about stack traces to analyzing a log file and building a spreadsheet. On the API task, both models got every graded field right. Claude just kept going: it added extra context about GitHub Actions tokens, Twilio’s queue window, OpenAI tier thresholds, and even a page it couldn’t reach. ChatGPT was faster, finishing in 1 minute 17 seconds; Claude took 1 minute 44 seconds and wrote far more.
The pattern held on the coding task. Both models passed all 16 hidden tests, but Claude again spent more time and tokens, and it flagged a judgment call about how to treat rounding at 0.5 seconds. On the log-analysis handoff, ChatGPT was quicker on the spreadsheet job, but Claude added a note about its percentile method and how the numbers would match in Google Sheets. The source’s bottom line is blunt: the answers were equally accurate, but ChatGPT was faster while Claude was more thorough.
My take — AI-written commentary, not fact-checked reporting
Speed wins demos; detail wins audits. That’s why the “same answers” line is a trap, and why the extra context Claude gave looks less like fluff than insurance. The industry keeps chasing feature parity, but the real split is whether a model stops at correct or bothers to explain itself.
Read more about this at: The New Stack
