Codex vs Devin
Direct comparison of Codex (from OpenAI) and Devin (from Cognition). Want a 3- or 4-way comparison? Open the multi-agent tool →
| Codex OpenAI | Devin Cognition | |
|---|---|---|
| Overview | ||
| Vendor | OpenAI | Cognition |
| Category | coding | coding |
| Release date | — | — |
| Source | proprietary | proprietary |
| License | — | — |
| Pricing | Included with ChatGPT Plus/Pro | $500/mo Team |
| Base models | — | |
| Capabilities | terminalfile_ioshell_execcloud_runtime | autonomouscloud_runtimemulti_sessionbrowser_use |
| Capability | ||
| gpt.buzz Index | 70.0 | 12.4 |
| SWE-bench Verified | 85.0% | 35.0% |
| SWE-bench Pro | 56.8% | — |
| Terminal-Bench 2.0 | 77.3% | — |
| GAIA | — | — |
| WebArena | — | — |
| Adoption | ||
| OpenRouter · daily | — | — |
| OpenRouter · cumulative | — | — |
| GitHub stars | — | — |
| Links | ||
| Links | ||
More on each
Full agent profiles — Codex / Devin
Or browse the full agent directory / composite leaderboard.