Jared PalmerDecision matchup
Kev 4Bvs.Kev 9B
Which model makes better decisions?
Kev 9B does better in independent tests.
Pick Kev 4B for
Pick Kev 9B for
Shared tests
Independent evaluators and leaderboards only, the newest shared snapshot per evaluator. With 95 % ranges a lead counts when they do not overlap; for hit rates, when it holds in a 95 % test. Without any uncertainty figure at most “slightly ahead”.
Jev Decision Index
Kev 9B aheadKev 9B is clearly ahead in 12 of 37 subtests and slightly in 6, Kev 4B in 3 clearly and slightly in 1. 15 are level.
POP909 · Decision Index 0.4.1
2,000 tasks, as of October 28, 2026
ChessBench · Decision Index 0.4.1
5,000 tasks, as of October 28, 2026
API-Bank · Decision Index 0.4.1
508 tasks, as of October 28, 2026
Show all 37 subtests
| Subtest | Kev 4B | Kev 9B | Verdict |
|---|---|---|---|
| POP909 · Decision Index 0.4.1 | 1.2 | 10.2 | Kev 9B ahead |
| ChessBench · Decision Index 0.4.1 | 8.9 | 12.0 | Kev 9B ahead |
| API-Bank · Decision Index 0.4.1 | 53.3 | 65.4 | Kev 9B ahead |
| GSM8K · Decision Index 0.4.1 | 46.5 | 54.4 | Kev 9B ahead |
| CRUXEval · Decision Index 0.4.1 | 48.1 | 55.4 | Kev 9B ahead |
| MMLU-Pro · Decision Index 0.4.1 | 50.7 | 56.5 | Kev 9B ahead |
| PhishNChips · Decision Index 0.4.1 | 55.3 | 61.6 | Kev 9B ahead |
| BPoMP · Decision Index 0.4.1 | 65.1 | 58.9 | Kev 4B ahead |
| When2Call · Decision Index 0.4.1 | 66.7 | 60.7 | Kev 4B ahead |
| cfcolor · Decision Index 0.4.1 | 57.9 | 53.0 | Kev 4B ahead |
| WinoGrande · Decision Index 0.4.1 | 70.2 | 75.1 | Kev 9B ahead |
| BBH · Decision Index 0.4.1 | 62.5 | 66.3 | Kev 9B ahead |
| Humicroedit · Decision Index 0.4.1 | 54.4 | 57.3 | Kev 9B ahead |
| HellaSwag · Decision Index 0.4.1 | 78.9 | 82.7 | Kev 9B ahead |
| BFCL · Decision Index 0.4.1 | 93.1 | 95.2 | Kev 9B ahead |
| ACOS · Decision Index 0.4.1 | 9.8 | 17.0 | Kev 9B slightly ahead |
| Amazon ESCI · Decision Index 0.4.1 | 39.7 | 49.1 | Kev 9B slightly ahead |
| ANLI · Decision Index 0.4.1 | 48.1 | 58.1 | Kev 9B slightly ahead |
| iSarcasmEval · Decision Index 0.4.1 | 46.8 | 43.2 | Kev 4B slightly ahead |
| VAST · Decision Index 0.4.1 | 52.4 | 56.7 | Kev 9B slightly ahead |
| ToolRet · Decision Index 0.4.1 | 61.0 | 63.8 | Kev 9B slightly ahead |
| RAGTruth · Decision Index 0.4.1 | 67.5 | 70.6 | Kev 9B slightly ahead |
| Home appliances · Decision Index 0.4.1 | 14.8 | 11.4 | level |
| New Yorker · Decision Index 0.4.1 | 52.3 | 56.6 | level |
| HLE · Decision Index 0.4.1 | 10.2 | 11.0 | level |
| SATA-Bench · Decision Index 0.4.1 | 28.8 | 30.7 | level |
| Habermas · Decision Index 0.4.1 | 42.0 | 40.0 | level |
| FinEntity · Decision Index 0.4.1 | 87.1 | 89.7 | level |
| CLINC150 · Decision Index 0.4.1 | 75.0 | 77.1 | level |
| BRIGHT · Decision Index 0.4.1 | 39.0 | 38.2 | level |
| HoVer · Decision Index 0.4.1 | 59.7 | 60.7 | level |
| ContractNLI · Decision Index 0.4.1 | 64.4 | 65.3 | level |
| NLI4CT · Decision Index 0.4.1 | 74.6 | 75.2 | level |
| CLadder · Decision Index 0.4.1 | 63.1 | 63.5 | level |
| BANKING77 · Decision Index 0.4.1 | 82.9 | 83.3 | level |
| MuSR · Decision Index 0.4.1 | 56.3 | 56.5 | level |
| GPQA Diamond · Decision Index 0.4.1 | 37.8 | 37.8 | level |
Source · latest October 28, 2026
AIMultiple decision-models evaluation
depends on the subtestKev 4B is clearly ahead in 0 of 4 subtests and slightly in 1, Kev 9B in 0 clearly and slightly in 1. 2 are level.
AIMultiple decision-task benchmark: Recall danger level
500 tasks, as of October 18, 2026
AIMultiple decision-task benchmark: Task done check
as of October 18, 2026
AIMultiple decision-task benchmark: Security flaw severity
500 tasks, as of October 18, 2026
Show all 4 subtests
| Subtest | Kev 4B | Kev 9B | Verdict |
|---|---|---|---|
| AIMultiple decision-task benchmark: Recall danger level | 61.10 | 56.70 | Kev 4B slightly ahead |
| AIMultiple decision-task benchmark: Task done check | 87.40 | 90.20 | Kev 9B slightly ahead |
| AIMultiple decision-task benchmark: Security flaw severity | 40.40 | 41.40 | level |
| AIMultiple decision-task benchmark: Research paper topic | 66.40 | 67.60 | level |
Source · latest October 18, 2026
Kevala decision benchmark report
Kev 9B slightly aheadKev 9B is clearly ahead in 0 of 3 subtests and slightly in 2, Kev 4B in none. 1 is level.
Kevala semif-perturbations108 decision accuracy
95 % ranges 81.2 to 98.8 and 90.7 to 100.0, 972 tasks, as of October 23, 2026
Kevala kevala-authored36 decision accuracy
95 % ranges 82.4 to 98.2 and 89.8 to 100.0, 324 tasks, as of October 23, 2026
Kevala semif-authored144 decision accuracy
95 % ranges 85.9 to 93.1 and 88.2 to 95.6, 1,296 tasks, as of October 23, 2026
Source · latest October 23, 2026
Duan et al., General Decision Models: Benchmarking and Insights Beyond Jev
Kev 9B aheadJEVal aida-testc accuracy
200 tasks, as of October 22, 2026
Source · latest October 22, 2026
Vendor figures
Figures from vendor reports that cover both models. They do not count toward the verdict.
Show 1 vendor figures
| Figure | Source | Kev 4B | Kev 9B |
|---|---|---|---|
| 14-public-dataset calibration error, per-dataset average (lower is better) | Jared Palmer, Introducing Kev | 0.10 | 0.10 |