Benchmarks
More
OverviewBenchmaxxingCompareChartsCost vs CapabilityProviders per ModelProvider explorerGatewaysEU & SovereignJev-class modelsAbout
JevBench by Benchmark Heaven · v1.4.2 · individual system

Cygnet (blockbrain, frozen Gemma-4-12B-it)

This detail uses the public, hash-checked v1.4.2 aggregate. Scores and ranks can change when a new release is published; the page preview remains name-only.

JevBench v1.4.2 score

61.8

Rank #4 of 89 ranked systems.

2.4 points behind Jev 1.13.0's 64.1.

Where it sits among the 89 ranked systems. The marked tick is Jev 1.13.0 (63.3).

Published axes

intelligence
49.5
calibration
74.9
speed
90.7
cost
52.8

Accuracy per tier, incl. sealed

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: Cygnet — system-one-open · Score 61.8 (#4)
  • B: decider-4b v2 — system-one-open · Score 64.1 (#1)

The four score axes

Radar: the four score axes, two systemsThe four score axes, Cygnet vs decider-4b v2. Intelligence: 49.5 vs 49.4; Calibration: 74.9 vs 75.0; Speed: 90.7 vs 92.9; Cost: 52.8 vs 60.9.50100Intelligence49.5 · 49.4Calibration74.9 · 75.0Speed90.7 · 92.9Cost52.8 · 60.9
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, Cygnet vs decider-4b v2. Easy: 100% vs 100%; Standard: 97% vs 97%; Judge: 95% vs 88%; Hard: 75% vs 67%; Sealed: 34% vs 35%.50100Easy100% · 100%Standard97% · 97%Judge95% · 88%Hard75% · 67%Sealed34% · 35%
Share correct per tier; Sealed = the 308 private decisions, aggregate only.

Hard tier by family (v1.2 topics)

Cygnet and decider-4b v2 have no published v1.2 hard-tier family breakdown.

Neither selected system has a published series for this view.

Share correct within each family of the 47 v1.2 hard-tier decisions (public and held-out).

Sealed set by family

Radar: sealed set by family, two systemsSealed set by family, Cygnet vs decider-4b v2. Ambiguous / abstain: 38% vs 43%; Judge: 54% vs 32%; Long policy: 23% vs 30%; Multi-hop: 32% vs 34%; Paraphrase: 64% vs 50%; Probability: 36% vs 36%; Safety judge: 44% vs 38%; Temporal / numeric: 18% vs 30%; Trade-off: 19% vs 35%; Trap / adversarial: 50% vs 33%.50100Ambiguous /abstain38% · 43%Judge54% · 32%Long policy23% · 30%Multi-hop32% · 34%Paraphrase64% · 50%Probability36% · 36%Safety judge44% · 38%Temporal /numeric18% · 30%Trade-off19% · 35%Trap /adversarial50% · 33%
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
shim MIT; weights Apache-2.0 with Google's Gemma Prohibited Use Policy
Cost evidence
estimated; the board’s row disclosure contains the published basis.
Endpoint condition
our evaluator-owned Lium GPU pod (RTX PRO 6000), offline read-only container, author's server on loopback
Published source
https://github.com/blockbrain-ai/cygnet-recipe

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