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

JEV Qwen3.5-9B Base NVFP4

Jev rebuild · by WilfLin

JevBench v1.4.2.2 score

37.7

Rank #28 of 91 ranked systems.

25.6 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
46.8
calibration
67.7
speed
93.3
cost
43.3

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.

The four score axes

Radar: the four score axes, two systemsThe four score axes, JEV Qwen3.5-9B Base NVFP4 vs Jev 1.13.0. Intelligence: 46.8 vs 53.1; Calibration: 67.7 vs 76.3; Speed: 93.3 vs 83.3; Cost: 43.3 vs 52.0.50100Intelligence46.8 · 53.1Calibration67.7 · 76.3Speed93.3 · 83.3Cost43.3 · 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, JEV Qwen3.5-9B Base NVFP4 vs Jev 1.13.0. Easy: 100% vs 100%; Standard: 92% vs 99%; Judge: 90% vs 95%; Hard: 59% vs 74%; Sealed: 30% vs 37%.50100Easy100% · 100%Standard92% · 99%Judge90% · 95%Hard59% · 74%Sealed30% · 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), JEV Qwen3.5-9B Base NVFP4 vs Jev 1.13.0. Ambiguous / abstain: 31% vs 43%; Judge: 54% vs 54%; Long policy: 28% vs 44%; Multi-hop: 44% vs 64%; Probability: 27% vs 63%; Temporal / numeric: 22% vs 28%; Trade-off: 47% vs 55%; Routing: 100% vs 100%; Trap / adversarial: 88% vs 83%; Paraphrase: 50% vs 64%; Safety judge: 50% vs 38%.50100Ambiguous /abstain31% · 43%Judge54% · 54%Long policy28% · 44%Multi-hop44% · 64%Probability27% · 63%Temporal /numeric22% · 28%Trade-off47% · 55%Routing100% · 100%Trap /adversarial88% · 83%Paraphrase50% · 64%Safety judge50% · 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, JEV Qwen3.5-9B Base NVFP4 vs Jev 1.13.0. Ambiguous / abstain: 16% vs 30%; Judge: 46% vs 34%; Long policy: 18% vs 28%; Multi-hop: 24% vs 45%; Paraphrase: 50% vs 64%; Probability: 14% vs 50%; Safety judge: 50% vs 38%; Temporal / numeric: 23% vs 29%; Trade-off: 38% vs 38%; Trap / adversarial: 67% vs 42%.50100Ambiguous /abstain16% · 30%Judge46% · 34%Long policy18% · 28%Multi-hop24% · 45%Paraphrase50% · 64%Probability14% · 50%Safety judge50% · 38%Temporal /numeric23% · 29%Trade-off38% · 38%Trap /adversarial67% · 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 entrant and upstream checkpoint
Cost evidence
estimated; the board’s row disclosure contains the published basis.
Endpoint condition
our GPU (lium.io RTX 5090 32 GB), local loopback HTTP, native FP4; serial
Note on this row
Byte-identical to upstream March-2026 NVFP4 checkpoint, predates JevBench v1.2, no task-specific training added. Had 3 preliminary scope/policy scoring errors corrected during review (pooled-ECE, full-run latency, cost estimate); score above is final corrected value.
Published source
https://huggingface.co/WIlfLin/JEV-Qwen3.5-9B-Base-NVFP4

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.