tokenjam 0.5.2 n=12 tasks · k=1 sample(s) · anthropic:claude-opus-4-7 → anthropic:claude-haiku-4-5
We ran the same 12 customer-support task(s) through both models and graded every answer with the same automated judge. The cheaper candidate (claude-haiku-4-5) cost 92% less than claude-opus-4-7, but scored 8 points lower on this suite — 100% of tasks passed before, 92% after.
No statistically significant quality drop was detected, so moving to the cheaper model is a reasonable way to cut cost.
| Task | Original | Cheaper model | Why — the judge’s reason |
|---|---|---|---|
| Refund window | pass | pass | The actual output accurately conveys the same factual information and meaning as graded 0.90 (needs 0.50) |
| Delivery delay | pass | pass | The actual output closely matches the expected output in terms of factual inform graded 0.90 (needs 0.50) |
| Wrong product | pass | pass | The actual output accurately reflects the factual information and intent of the graded 0.90 (needs 0.50) |
| Password reset | pass | pass | The actual output closely aligns with the expected output in terms of factual in graded 0.90 (needs 0.50) |
| Subscription cancel | pass | pass | The actual output is factually accurate and semantically equivalent to the expec graded 0.90 (needs 0.50) |
| Payment failed | pass | pass | The actual output is factually accurate and semantically equivalent to the expec graded 0.80 (needs 0.50) |
| Invoice request | pass | fail | The actual output does not align with the expected output in terms of factual ac graded 0.10 (needs 0.50) |
| Account locked | pass | pass | The actual output is factually accurate and semantically equivalent to the expec graded 0.90 (needs 0.50) |
| Shipping status | pass | pass | The actual output closely aligns with the expected output in terms of factual in graded 0.90 (needs 0.50) |
| Feature request | pass | pass | The actual output matches the expected output exactly in terms of factual inform graded 1.00 (needs 0.50) |
| Bug report | pass | pass | The actual output matches the expected output exactly in terms of factual accura graded 1.00 (needs 0.50) |
| Escalation | pass | pass | The actual output is factually accurate and semantically equivalent to the expec graded 0.90 (needs 0.50) |
McNemar’s test asks whether the difference between the two models is bigger than chance: a p-value above 0.05 means the change is not statistically significant. The 95% CI on the pass-rate delta is the range the true difference is likely in — if it crosses zero, the direction isn’t certain.