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 email-assistant task(s) through both models and graded every answer with the same automated judge. The cheaper candidate (claude-haiku-4-5) cost 85% less than claude-opus-4-7, and scored about the same on this suite — 100% of tasks passed before, 100% 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 |
|---|---|---|---|
| Reply reschedule | pass | pass | The actual output is professional and aligns well with the expected output. It i graded 0.90 (needs 0.50) |
| Reply refund firm | pass | pass | The actual output aligns well with the expected output in terms of factual accur graded 0.90 (needs 0.50) |
| Summarize thread | pass | pass | The actual output closely aligns with the expected output, accurately reflecting graded 0.90 (needs 0.50) |
| Triage priority | pass | pass | The actual output is factually accurate and semantically equivalent to the expec graded 0.90 (needs 0.50) |
| Triage spam | pass | pass | The actual output closely aligns with the expected output, accurately identifyin graded 0.90 (needs 0.50) |
| Extract meeting | pass | pass | The actual output accurately reflects the factual information from the expected graded 0.80 (needs 0.50) |
| Extract actions | pass | pass | The actual output closely matches the expected output in terms of factual inform graded 0.80 (needs 0.50) |
| Followup noreply | pass | pass | The actual output closely aligns with the expected output in terms of semantic m graded 0.90 (needs 0.50) |
| Reply decline meeting | pass | pass | The actual output contains all the factual information present in the expected o graded 0.90 (needs 0.50) |
| Summarize exec | pass | pass | The actual output closely aligns with the expected output in terms of factual in graded 0.90 (needs 0.50) |
| Triage route | pass | pass | The actual output aligns well with the expected output in terms of factual accur graded 0.90 (needs 0.50) |
| Reply sensitive data | pass | pass | The actual output accurately conveys the key points of the expected output, such graded 0.80 (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.