TypeSafe AIMapikaDecision matchup
Jevvs.Decider 4B
Which model makes better decisions?
Jev does better in independent tests, while Decider 4B is available with open weights.
Independent testsDecideBench
Jevahead1 of 1
Decider 4B0 of 1
Popularityout of 100
Jev43
Decider 4B43
Weights and access
Jevn/a
Decider 4Bopenopen, local
SizeParameters
Jevn/a
Decider 4B4 B
Pick Jev for
DecideBench
Pick Decider 4B for
open weightsruns locally
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”.
DecideBench
Jev aheadJev is clearly ahead in 6 of 10 subtests, Decider 4B in none. 4 are level.
DecideBench pair accuracy
Jev96.0 %
Decider 4B78.5 %
200 tasks, as of October 18, 2026
DecideBench action_review accuracy
Jev100.0 %
Decider 4B82.0 %
50 tasks, as of October 18, 2026
DecideBench returns_policy accuracy
Jev96.0 %
Decider 4B82.0 %
50 tasks, as of October 18, 2026
Show all 10 subtests
| Subtest | Jev | Decider 4B | Verdict |
|---|---|---|---|
| DecideBench pair accuracy | 96.0 % | 78.5 % | Jev ahead |
| DecideBench action_review accuracy | 100.0 % | 82.0 % | Jev ahead |
| DecideBench returns_policy accuracy | 96.0 % | 82.0 % | Jev ahead |
| DecideBench content_moderation accuracy | 98.0 % | 86.0 % | Jev ahead |
| DecideBench ticket_triage accuracy | 98.0 % | 86.0 % | Jev ahead |
| DecideBench overall accuracy | 98.0 % | 88.5 % | Jev ahead |
| DecideBench review_sentiment accuracy | 98.0 % | 88.0 % | level |
| DecideBench claim_support accuracy | 96.0 % | 90.0 % | level |
| DecideBench agent_routing accuracy | 98.0 % | 96.0 % | level |
| DecideBench support_intent accuracy | 100.0 % | 98.0 % | level |
Source · latest October 18, 2026
All data side by side
| Jev | Decider 4B | |
|---|---|---|
| Kind | Standalone | Standalone |
| Developer | TypeSafe AI | Mapika |
| Status | Early access | Open source |
| Released | September 15, 2026 | n/a |
| Weights | n/a | open |
| Runs locally | no | yes |
| Parameters | n/a | 4 B |
| License | n/a | Apache-2.0 |
| Popularity (out of 100) | 43 | 43 |