Benchmarks
More
OverviewBenchmaxxingCompareChartsCost vs CapabilityProviders per ModelProvider explorerGatewaysEU & SovereignJev-class modelsImage JevBenchAbout
JevBench by Benchmark Heaven · v1.4.2.2 · individual system

OpenDecision (ModernBERT-large zero-shot)

Zero-shot classifier · by Deepan Wadhwa

JevBench v1.4.2.2 score

21.6

Rank #58 of 91 ranked systems.

41.7 points behind Jev 1.13.0's 63.3.

Where it sits among the 91 ranked systems. The marked tick is Jev 1.13.0 (63.3).

Published axes

intelligence
31.8
calibration
57.1
speed
79.9
cost
75.3

Against Jev 1.13.0

Four radars compare this fixed pair across the score axes, accuracy per tier, and accuracy by family on the hard tier and sealed set. Further out is better on every spoke.

  • A: OpenDecision — Zero-shot classifier · Score 21.6 (#58)
  • B: Jev 1.13.0 — Jev (TypeSafe, closed) · Score 63.3 (#4)

The four score axes

Radar: the four score axes, two systemsThe four score axes, OpenDecision vs Jev 1.13.0. Intelligence: 31.8 vs 53.1; Calibration: 57.1 vs 76.3; Speed: 79.9 vs 83.3; Cost: 75.3 vs 52.0.50100Intelligence31.8 · 53.1Calibration57.1 · 76.3Speed79.9 · 83.3Cost75.3 · 52.0
0–100, the values in the table. A label-only system has no calibration (counted as 0).

Accuracy per tier, incl. sealed

Radar: accuracy per tier, incl. sealed, two systemsAccuracy per tier, incl. sealed, OpenDecision vs Jev 1.13.0. Easy: 88% vs 100%; Standard: 63% vs 99%; Judge: 71% vs 95%; Hard: 33% vs 74%; Sealed: 26% vs 37%.50100Easy88% · 100%Standard63% · 99%Judge71% · 95%Hard33% · 74%Sealed26% · 37%
Share correct per tier; Sealed = the 308 private decisions, aggregate only.

Current question set by family (hard + sealed)

Radar: current question set by family (hard + sealed), two systemsCurrent question set by family (hard + sealed), OpenDecision vs Jev 1.13.0. Ambiguous / abstain: 27% vs 43%; Judge: 47% vs 54%; Long policy: 22% vs 44%; Multi-hop: 11% vs 64%; Probability: 31% vs 63%; Temporal / numeric: 20% vs 28%; Trade-off: 29% vs 55%; Routing: 10% vs 100%; Trap / adversarial: 65% vs 83%; Paraphrase: 21% vs 64%; Safety judge: 31% vs 38%.50100Ambiguous /abstain27% · 43%Judge47% · 54%Long policy22% · 44%Multi-hop11% · 64%Probability31% · 63%Temporal /numeric20% · 28%Trade-off29% · 55%Routing10% · 100%Trap /adversarial65% · 83%Paraphrase21% · 64%Safety judge31% · 38%
Share correct per family across the 220 hard-tier decisions (public and held out) and the 308 sealed decisions of v1.4, pooled; Routing is hard-tier only, Paraphrase and Safety judge sealed only.

Sealed set by family

Radar: sealed set by family, two systemsSealed set by family, OpenDecision vs Jev 1.13.0. Ambiguous / abstain: 32% vs 30%; Judge: 46% vs 34%; Long policy: 15% vs 28%; Multi-hop: 8% vs 45%; Paraphrase: 21% vs 64%; Probability: 25% vs 50%; Safety judge: 31% vs 38%; Temporal / numeric: 18% vs 29%; Trade-off: 31% vs 38%; Trap / adversarial: 50% vs 42%.50100Ambiguous /abstain32% · 30%Judge46% · 34%Long policy15% · 28%Multi-hop8% · 45%Paraphrase21% · 64%Probability25% · 50%Safety judge31% · 38%Temporal /numeric18% · 29%Trade-off31% · 38%Trap /adversarial50% · 42%
Share correct within each sealed family — system-level aggregates; the items stay private.

Availability and evidence

Openness
Code and weights marked open in the published row
License note
Apache-2.0
Cost evidence
estimated; the board’s row disclosure contains the published basis.
Endpoint condition
our RunPod GPU (H100 NVL 96 GB, Canada), reached over the internet from Germany
Note on this row
A zero-shot NLI classifier behind a TypeSafe-compatible server, not a trained decision model: it scores each option as an entailment hypothesis with ModernBERT-large-zeroshot-v2.0. Its choice path runs several NLI passes over the same state, which the reported token count does not include, so a per-token hosted price would be higher than the estimate here. Pre-registered for our CPU in v1.2.7, run on our GPU because the CPU was far too slow.
Published source
https://github.com/deepanwadhwa/OpenDecision

From the public v1.4.2.2 aggregate. Scores and ranks can change when a new release is published.

Read the full board and published method. The overall score is a composite, not raw accuracy.