Real token-savings benchmarks for Collar and dotdog. Submit yours with ./dogbench --json.
| Date | User | Collar | Claude Code | Codex | Savings |
|---|---|---|---|---|---|
| — | Your name here | — | — | — | — |
- Run:
./dogbench --json > results.json - Open a PR adding your row to the table above
- Include your
results.json
Or paste the output directly into an issue.
| Harness | Engine | Task | Tokens | Cost |
|---|---|---|---|---|
| collar (DAG-first) | deepseek-v4-pro | CSV sales report | ~1,800 | $0.0018 |
| Claude Code | claude-sonnet-4 | CSV sales report | ~4,500 | $0.0284 |
| OpenAI Codex | gpt-5 | CSV sales report | ~3,200 | $0.0221 |
Estimates from internal testing. Submit your own results above.