GPT-6 Luna (default medium reasoning effort)v1.7 · 10 s rule
Data: JevBench v1.8.7, published 2026-11-03 · method v1.7 (10 s rule), re-scored from the recorded per-decision times
Capability 98.4 · API offering, ranked on the API leaderboard · outside the Jev-class caps (cost 3.5× Jev (v1.5 reference); latency 3.1× Jev (v1.5 reference)), below the Capability ranking
Measured 2026-10-06 on A4 ∪ P, 600 items, equated (+2.57 I / +3.35 C); round score-a4-2.
proprietary API
intelligence
96.9
calibration
99.9
speed
71.4
cost
38.4
Composite (secondary)
39.1 · Composite rank #32 of 48 on the API leaderboard
Cost and measurement conditions
$0.113 per 1,000 decisions (public tariff)
measured: token usage of the 6 Oct 2026 A4 u P run x tariff USD 0.1/M input, 0.5/M output (spec gpt-6-luna.json), per 1,000 decisions on the v1.6.1 common cost basis (591 of 600 items; the 9 A4 u P items outside the Jev reference's accepted input range are excluded, cost-basis-a4.json)
p50 latency: 1.9 s
none (API)
Measured latency over 600 decisions: p50 1.932 s · p95 3.765 s
5 decisions over 10 s (1 of them answered without error, counted as wrong).
A decision that takes longer than 10 seconds (measured time) counts as a wrong answer and as zero calibration for that item: each type's calibration is multiplied by the share of its decisions answered within 10 s. The limit uses the measured time (raw). The displayed adjusted time for self-hosted rows (2 × + 0.15 s) is an assumption and is not used for the limit.
JevBench v1.6.43 values (method v1.6, before the 10 s rule)
Capability 98.44 · Composite 39.16 · rank #84 on the all-systems board · #30 API
intelligence 96.88 · calibration 100.00 · speed 71.77 · cost 38.41
v1.6 speed basis (124 items): p50 1.824 s · p95 3.646 s
These figures are kept for comparison. The current figures above are re-scored under the 10 s rule from the recorded per-decision times and replace them.
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 | 99.2 | 99.2 |
| noul | 97.1 | 93.4 |
| score | 89.7 | 87.2 |
Capability by subject topic
| Rules, policy & law | 98.9 | n=1133 |
|---|---|---|
| Coding & software | 90.5 | n=202 |
| Math & numbers | 91.9 | n=269 |
| Finance & commerce | 95.7 | n=178 |
| Support & operations | 98.8 | n=109 |
| Everyday language | 99.4 | n=130 |
| Safety & security | 99.8 | n=84 |
Use cases (TypeSafe categories)
| Model routing | 93.0 | n=267 |
|---|---|---|
| Legal & compliance | 99.3 | n=269 |
| Customer support | 99.5 | n=156 |
| E-commerce | 96.6 | n=88 |
| Insurance claims | 95.2 | n=92 |
| Risk assessment | 100.0 | n=88 |
| Financial crime | 98.7 | n=76 |
| Feature extraction | 96.7 | n=90 |
| Lead generation | 97.7 | n=86 |
| Recruiting | 100.0 | n=88 |
| Knowledge graphs | 97.2 | n=82 |
| LLM guardrails | 99.9 | n=82 |
| Moderation | 98.4 | n=69 |
| Code linting | 95.3 | n=81 |
| Search & retrieval | 99.8 | n=84 |
| Science | 99.9 | n=82 |
| Advertising | 98.0 | n=85 |
| Gaming | 87.9 | n=79 |
| Demand forecasting | 100.0 | 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 98.4 · composite 39.1 · cost $0.113 per 1,000 decisions (public list price)
Last measured 2026-10-06, measurement release v1.6.1. Older method and scale; the current figures above replace it.