system-one-open (Gemma 4 E2B LoRA on an L4)
Data: JevBench v1.6.1, published 2026-10-06
Capability 51.9 · API offering, ranked on the API leaderboard · Capability rank #11 of 14 within the Jev-class caps
Measured 2026-10-06 on A4 ∪ P, 600 items, equated (+2.57 I / +3.35 C); round score-a4-2.
MIT (repository LICENSE; Gemma weights keep Google’s terms)
intelligence
27.9
calibration
75.9
speed
79.3
cost
68.3
Composite (secondary)
16.4 · Composite rank #20 of 27 on the API leaderboard
Cost and measurement conditions
$0.011 per 1,000 decisions (estimated)
v1.5.4 published cost per 1,000 decisions carried (pricing rules unchanged; v1.6 item lengths differ): ESTIMATE: v1.5 measured-input proxy; no exact base-model floor applies
p50 latency: 1.0 s
x2 (demo)
Breakdowns (not part of the score)
Chance-corrected competence (0 = chance, 100 = perfect) on A4 ∪ P (600 items: 300 open + 300 sealed; fewer sealed items than the full set, so wider uncertainty). Raw values, not equated. Category cells use A4+P+L1+L2+L3; per-type/tier cells use the original basis above. Radars against Jev.
| Request type | Open | Sealed |
|---|---|---|
| choice | 56.8 | 47.8 |
| noul | 3.8 | 9.8 |
| score | 23.1 | 10.9 |
Capability by subject topic
| Rules, policy & law | 8.0 | n=1133 |
|---|---|---|
| Coding & software | 25.8 | n=202 |
| Math & numbers | 0.0 | n=269 |
| Finance & commerce | 13.1 | n=178 |
| Support & operations | 24.7 | n=109 |
| Everyday language | 48.3 | n=130 |
| Safety & security | 0.0 | n=84 |
Use cases (TypeSafe categories)
| Model routing | 43.6 | n=267 |
|---|---|---|
| Legal & compliance | 26.5 | n=269 |
| Customer support | 15.4 | n=156 |
| E-commerce | 8.8 | n=88 |
| Insurance claims | 0.0 | n=92 |
| Risk assessment | 7.0 | n=88 |
| Financial crime | 0.0 | n=76 |
| Feature extraction | 0.0 | n=90 |
| Lead generation | 26.2 | n=86 |
| Recruiting | 1.1 | n=88 |
| Knowledge graphs | 2.1 | n=82 |
| LLM guardrails | 22.6 | n=82 |
| Moderation | 0.0 | n=69 |
| Code linting | 0.0 | n=81 |
| Search & retrieval | 9.8 | n=84 |
| Science | 0.0 | n=82 |
| Advertising | 0.0 | n=85 |
| Gaming | 4.4 | n=79 |
| Demand forecasting | 0.0 | n=72 |
Cells need at least 15 answered items; † fewer than 30 (indicative only, listed below the radar in the compare view).
Previous measurement (v1.5.x, not comparable)
Capability 56.6 · composite 42.4 · cost $0.011 per 1,000 decisions (estimated)
Last measured 2026-09-28, measurement release v1.5.0. Older method and scale; the current figures above replace it.