Mistral Small 3.2 24B Instruct (Venice) vs Qwen2.5 7B (local)
On the 5 use cases both models ran, Mistral Small 3.2 24B Instruct (Venice) and Qwen2.5 7B (local) each won 0, with 5 tied (better outcome first, then lower cost).
Pass rate: Mistral Small 3.2 24B Instruct (Venice) 0% (5 runs) vs Qwen2.5 7B (local) 0% (5 runs). Average cost per run: $0.01 vs $0 (local).
| Metric | Mistral Small 3.2 24B Instruct (Venice) | Qwen2.5 7B (local) |
|---|---|---|
| Runs on shared use cases | 5 | 5 |
| Pass rate | 0% (0/5) | 0% (0/5) |
| Avg cost per run | $0.01 | $0 (local) |
| Avg wall time | 4m | 58s |
| Step-ins per run | 2.00 | 1.20 |
| Use cases won | 0 | 0 |
Use case by use case
- Audit expense claims with per-day caps and unverified exception notes: tie (pass 0% vs 0%)
- Build a vendor payout batch when one vendor's amount units are unknown: tie (pass 0% vs 0%)
- Finish a spoken in-car request unattended: tie (pass 0% vs 0%)
- Plan a picking robot's shortest route on a grid with an unconfirmed hazard: tie (pass 0% vs 0%)
- Replay a 600-event inventory ledger with an undefined return policy and write-off sign-off: tie (pass 0% vs 0%)
Only use cases both models ran are compared; all settings of each model are grouped. How we score.