Intelligence
How often the system answers correctly.
Model benchmark · JevBench v1.3.0
Jev models, compared on the same benchmark.
Explore published results for System One decision models. Compare overall scores, decision quality, calibration, speed, and cost to help choose a model for your workload.
Selected ranking
Jev model benchmark scores
A selection of systems from the public board. The composite score weights intelligence, calibration, speed, and cost equally; see the source for the full ranking and interactive comparisons.
| Rank | Model / system | Score | Intelligence | Calibration | Speed | Cost / 1,000 |
|---|---|---|---|---|---|---|
| #1 | Jev 1.13.0TypeSafe AI | 74.4 | 85.7 | 82.7 | 83.3 | $0.040Published price |
| #2 | SemIfQwen3.5-4B · open rebuild | 73.1 | 79.0 | 72.6 | 83.7 | ~$0.022Estimated |
| #3 | djevMaisa · DiffusionGemma | 73.0 | 82.7 | 65.4 | 91.4 | ~$0.026Announced price |
| #11 | OpenJevDiffusionGemma 26B-A4B | 66.4 | 79.2 | 64.8 | 83.2 | ~$0.066Estimated |
| #33 | LayaModernBERT-large | 54.4 | 45.8 | 62.5 | 71.1 | ~$0.003Estimated |
Costs are USD per 1,000 complete decisions. Estimates and announced but uncharged prices are labelled; these third-party benchmark figures are not Jev AI Model API prices or a promise of cost for your workload.
Source: Benchmark Heaven · JevBench v1.3.0 · Snapshot: September 21, 2026
How to read the scores
What does the benchmark tell you?
The overall score is a quick view of relative performance on one shared task set. Before production, validate accuracy, confidence thresholds, latency, and cost on your own examples.
Calibration
Whether confidence reflects observed outcomes.
Speed & cost
Latency and the price of complete decisions.
Try Jev AI Model yourself
Test a decision on your own example.
The online playground is free after sign in. Buy credits for API calls.
Jev is a model released by TypeSafe AI. Jev AI Model is an independent online playground and API service. This page cites a third-party benchmark for reference and is not an endorsement by the model provider or benchmark publisher.