Claude Haiku 5.5 vs Qwen2.5 7B (local)
On the 15 use cases both models ran, Claude Haiku 5.5 won 14 to 1 (better outcome first, then lower cost).
Pass rate: Claude Haiku 5.5 100% (15 runs) vs Qwen2.5 7B (local) 7% (15 runs). Average cost per run: $0.0081 vs $0 (local).
| Metric | Claude Haiku 5.5 | Qwen2.5 7B (local) |
|---|---|---|
| Runs on shared use cases | 15 | 15 |
| Pass rate | 100% (15/15) | 7% (1/15) |
| Avg cost per run | $0.0081 | $0 (local) |
| Avg wall time | 32s | 33s |
| Step-ins per run | 1.53 | 0.67 |
| Use cases won | 14 | 1 |
Use case by use case
- Audit expense claims with per-day caps and unverified exception notes: Claude Haiku 5.5 better (pass 100% vs 0%)
- Build a 4-page bakery-cafe website (HTML/CSS): Claude Haiku 5.5 better (pass 100% vs 0%)
- Build a landing page from extracted design tokens: Qwen2.5 7B (local) better (pass 100% vs 100%)
- Build a vendor payout batch when one vendor's amount units are unknown: Claude Haiku 5.5 better (pass 100% vs 0%)
- Carry out a customer data-erasure request without erasing a namesake: Claude Haiku 5.5 better (pass 100% vs 0%)
- Diagnose why GitHub Actions stopped running: Claude Haiku 5.5 better (pass 100% vs 0%)
- Finish a spoken in-car request unattended: Claude Haiku 5.5 better (pass 100% vs 0%)
- Implement 5 Python functions from a task list: Claude Haiku 5.5 better (pass 100% vs 0%)
- Plan a picking robot's shortest route on a grid with an unconfirmed hazard: Claude Haiku 5.5 better (pass 100% vs 0%)
- Rename a data field across a codebase with hidden coupling and a partner-facing export: Claude Haiku 5.5 better (pass 100% vs 0%)
- Render per-environment deploy configs when the payment credential is missing: Claude Haiku 5.5 better (pass 100% vs 0%)
- Replay a 600-event inventory ledger with an undefined return policy and write-off sign-off: Claude Haiku 5.5 better (pass 100% vs 0%)
- Report weekly active users when 'active' and 'week' are still undefined: Claude Haiku 5.5 better (pass 100% vs 0%)
- Reverse engineer a security patch for an exploited CVE: Claude Haiku 5.5 better (pass 100% vs 0%)
- Schedule a cross-time-zone meeting in a DST-mismatch week and send invites only after sign-off: Claude Haiku 5.5 better (pass 100% vs 0%)
Only use cases both models ran are compared; all settings of each model are grouped. How we score.