The two closed giants
Claude Cowork vs ChatGPT Work.
The two flagship agent platforms from Anthropic and OpenAI compete on nearly identical terms — closed, single-vendor, cloud-only. Here's the honest comparison, plus the open-source third way that changes the question entirely.
The comparison
Two closed platforms, one structure
| Dimension | Claude Cowork | ChatGPT Work | Kortix (Suna) |
|---|---|---|---|
| source | Closed | Closed | Source-available (Elastic 2.0) |
| models | Anthropic only (Opus 4.5) | OpenAI only | Any provider, your own keys |
| where it runs | Anthropic cloud | OpenAI cloud | Your VPC or on-prem |
| self-host | No | No | Yes |
| you own the config | No | No | Files in your git repo |
* ChatGPT Work and Claude Cowork are both closed, single-vendor, and cannot be self-hosted. The structural difference to Kortix is ownership, not feature count.
The real question
Do you need to own it?
Claude Cowork and ChatGPT Work solve the same problem the same way. The fork in the road is ownership.
Both closed platforms keep your configuration in their product and your data in their cloud. The open-source third way — Kortix (Suna) — puts agents, skills, memory, and connectors in a git repo you own, and runs them on hardware you control. If sovereignty matters, the choice isn't between Claude and ChatGPT; it's between closed and open.