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

Jobe Qwen3.5-4B (frozen)

Native-logit decision engine · by MantisShrimpdev

JevBench v1.4.2.2 score

46.9

Rank #14 of 91 ranked systems.

16.4 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
44.1
calibration
66.1
speed
85.6
cost
59.5

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: Jobe Qwen3.5-4B — Native-logit decision engine · Score 46.9 (#14)
  • 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, Jobe Qwen3.5-4B vs Jev 1.13.0. Intelligence: 44.1 vs 53.1; Calibration: 66.1 vs 76.3; Speed: 85.6 vs 83.3; Cost: 59.5 vs 52.0.50100Intelligence44.1 · 53.1Calibration66.1 · 76.3Speed85.6 · 83.3Cost59.5 · 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, Jobe Qwen3.5-4B vs Jev 1.13.0. Easy: 100% vs 100%; Standard: 98% vs 99%; Judge: 97% vs 95%; Hard: 59% vs 74%; Sealed: 26% vs 37%.50100Easy100% · 100%Standard98% · 99%Judge97% · 95%Hard59% · 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), Jobe Qwen3.5-4B vs Jev 1.13.0. Ambiguous / abstain: 25% vs 43%; Judge: 55% vs 54%; Long policy: 28% vs 44%; Multi-hop: 40% vs 64%; Probability: 29% vs 63%; Temporal / numeric: 22% vs 28%; Trade-off: 42% vs 55%; Routing: 100% vs 100%; Trap / adversarial: 83% vs 83%; Paraphrase: 36% vs 64%; Safety judge: 38% vs 38%.50100Ambiguous /abstain25% · 43%Judge55% · 54%Long policy28% · 44%Multi-hop40% · 64%Probability29% · 63%Temporal /numeric22% · 28%Trade-off42% · 55%Routing100% · 100%Trap /adversarial83% · 83%Paraphrase36% · 64%Safety judge38% · 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, Jobe Qwen3.5-4B vs Jev 1.13.0. Ambiguous / abstain: 14% vs 30%; Judge: 37% vs 34%; Long policy: 18% vs 28%; Multi-hop: 18% vs 45%; Paraphrase: 36% vs 64%; Probability: 21% vs 50%; Safety judge: 38% vs 38%; Temporal / numeric: 25% vs 29%; Trade-off: 27% vs 38%; Trap / adversarial: 58% vs 42%.50100Ambiguous /abstain14% · 30%Judge37% · 34%Long policy18% · 28%Multi-hop18% · 45%Paraphrase36% · 64%Probability21% · 50%Safety judge38% · 38%Temporal /numeric25% · 29%Trade-off27% · 38%Trap /adversarial58% · 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
MIT code; Apache-2.0 weights
Cost evidence
estimated; the board’s row disclosure contains the published basis.
Endpoint condition
our GPU (lium.io RTX A6000 48 GB), in-process, serial
Note on this row
No trained weights/LoRA/calibration fit; release explicitly rejects fitted temperature/order averaging. Scan of 37 files vs 231 public tasks: 0 matches.
Published source
https://github.com/MantisShrimpdev/jobe

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.