Pyyan / Compare / Claude Cowork vs Claude Code vs Codex vs Dots
Claude Cowork vs Claude Code vs Codex vs Dots
4 of 5Agentic Tools · verified 1 Oct 2026
| Specification | Claude Cowork | Claude Code | Codex | Dots |
|---|---|---|---|---|
| Summary | Knowledge-work agent for business users. | Terminal agent with per-subagent model control. | Holds the Terminal-Bench record. | An agent with its own computer, still running after you close the tab. |
| Surface | Web | CLI + IDE + web | CLI + cloud | ChatGPT + Slack + Teams |
| Price | from $20/mo | — | — | from $100/mo |
| Subagents | — | Yes | No | Yes |
| Category | Agentic Tools | Agentic Tools | Agentic Tools | Agentic Tools |
| Official | Anthropic ↗ | Anthropic ↗ | OpenAI ↗ | OpenAI ↗ |
Highlighted rows are where these differ.
Claude Code
- Built as an agent, not an assistant
- Discovers and reads files unprompted
- Strongest on multi-file reasoning
Best for autonomous work.
Dots
- Each dot gets its own cloud computer and browser, on GPT-6 Astra
- Reaches the 4,000 or so apps already connected to ChatGPT, and delegates to sub-agents
- Nothing on Free, Go or Plus, and not for Pro users in the EEA, Switzerland or the UK
Best for work that continues without you.