Microsoft-Decision-1 (Azure Foundry)
Data: JevBench v1.6.1, published 2026-10-06
Capability 70.8 · API offering, ranked on the API leaderboard · Capability rank #7 of 14 within the Jev-class caps
Measured 2026-10-10 on the full v1.6.1 set (1500 items), scored like the other full-set API rows, not equated.
Licence not stated
intelligence
57.2
calibration
84.4
speed
81.5
cost
61.5
Composite (secondary)
69.1 · Composite rank #6 of 27 on the API leaderboard
Cost and measurement conditions
$0.019 per 1,000 decisions (public tariff)
Advertised Azure Foundry tariff .042/M input, output free; genuine native usage, billed invoice unknown
p50 latency: 0.5 s
none (API)
Breakdowns (not part of the score)
Chance-corrected competence (0 = chance, 100 = perfect) on the full set (1500 items). Raw values, not equated. Category cells use S1200+P300+L1+L2+L3; full native Foundry union; per-type/tier cells use the original basis above. Radars against Jev.
| Request type | Open | Sealed |
|---|---|---|
| choice | 82.4 | 65.8 |
| noul | 51.5 | 37.6 |
| score | 62.7 | 45.6 |
Capability by subject topic
| Rules, policy & law | 53.2 | n=1600 |
|---|---|---|
| Coding & software | 57.4 | n=277 |
| Math & numbers | 20.7 | n=459 |
| Finance & commerce | 56.3 | n=236 |
| Support & operations | 56.3 | n=167 |
| Everyday language | 75.1 | n=152 |
| Safety & security | 53.6 | n=114 |
Use cases (TypeSafe categories)
| Model routing | 64.3 | n=400 |
|---|---|---|
| Legal & compliance | 63.4 | n=471 |
| Customer support | 60.4 | n=249 |
| E-commerce | 52.6 | n=114 |
| Insurance claims | 42.8 | n=107 |
| Risk assessment | 51.2 | n=126 |
| Financial crime | 29.0 | n=116 |
| Feature extraction | 36.9 | n=96 |
| Lead generation | 63.9 | n=114 |
| Recruiting | 29.7 | n=98 |
| Knowledge graphs | 35.5 | n=98 |
| LLM guardrails | 69.7 | n=101 |
| Moderation | 67.2 | n=98 |
| Code linting | 19.8 | n=96 |
| Search & retrieval | 85.3 | n=97 |
| Science | 36.9 | n=97 |
| Advertising | 37.0 | n=107 |
| Gaming | 25.1 | n=90 |
| Demand forecasting | 0.0 | n=88 |
Cells need at least 15 answered items; † fewer than 30 (indicative only, listed below the radar in the compare view).