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

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

system-one-open · by blockbrain

JevBench v1.4.2.2 score

61.8

Rank #6 of 91 ranked systems.

1.5 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
49.5
calibration
74.9
speed
90.7
cost
52.8

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: Cygnet — system-one-open · Score 61.8 (#6)
  • 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, Cygnet vs Jev 1.13.0. Intelligence: 49.5 vs 53.1; Calibration: 74.9 vs 76.3; Speed: 90.7 vs 83.3; Cost: 52.8 vs 52.0.50100Intelligence49.5 · 53.1Calibration74.9 · 76.3Speed90.7 · 83.3Cost52.8 · 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, Cygnet vs Jev 1.13.0. Easy: 100% vs 100%; Standard: 97% vs 99%; Judge: 95% vs 95%; Hard: 75% vs 74%; Sealed: 34% vs 37%.50100Easy100% · 100%Standard97% · 99%Judge95% · 95%Hard75% · 74%Sealed34% · 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), Cygnet vs Jev 1.13.0. Ambiguous / abstain: 53% vs 43%; Judge: 69% vs 54%; Long policy: 44% vs 44%; Multi-hop: 53% vs 64%; Probability: 50% vs 63%; Temporal / numeric: 23% vs 28%; Trade-off: 39% vs 55%; Routing: 100% vs 100%; Trap / adversarial: 85% vs 83%; Paraphrase: 64% vs 64%; Safety judge: 44% vs 38%.50100Ambiguous /abstain53% · 43%Judge69% · 54%Long policy44% · 44%Multi-hop53% · 64%Probability50% · 63%Temporal /numeric23% · 28%Trade-off39% · 55%Routing100% · 100%Trap /adversarial85% · 83%Paraphrase64% · 64%Safety judge44% · 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, Cygnet vs Jev 1.13.0. Ambiguous / abstain: 38% vs 30%; Judge: 54% vs 34%; Long policy: 23% vs 28%; Multi-hop: 32% vs 45%; Paraphrase: 64% vs 64%; Probability: 36% vs 50%; Safety judge: 44% vs 38%; Temporal / numeric: 18% vs 29%; Trade-off: 19% vs 38%; Trap / adversarial: 50% vs 42%.50100Ambiguous /abstain38% · 30%Judge54% · 34%Long policy23% · 28%Multi-hop32% · 45%Paraphrase64% · 64%Probability36% · 50%Safety judge44% · 38%Temporal /numeric18% · 29%Trade-off19% · 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
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
Note on this row
Blockbrain-ai/cygnet-recipe 81974de: frozen google/gemma-4-12B-it rev 707f0a3b8a3c7ad586ed01e27eafbad8a27dd0f7 on unmodified vLLM 0.30.0 (image vllm/vllm-openai:v0.30.0@sha256:8a69ffad…), the author's shim: options as letters in the benchmark's label order, logits masked to the option letters, one calibration temperature T=3.4 fitted on the author's own items. The shim sends its own system prompt and renders structured state with json indent=1. Over-context/over-26-option inputs are HTTP 422 (a wrong answer). Offline self-hosted inference of all 842 decisions (534 frozen v1.2 + 308 sealed v1.4) in a network-disabled, read-only container on an evaluator-owned Lium RTX PRO 6000 Blackwell 96 GB pod, through JevBench's unchanged typesafe adapter against the author's own server on loopback; no operator endpoint; no golds were exposed. Latency is the serial request wall time on the standard+judge items with the self-hosted adjustment. Cost is a labelled estimate: OpenRouter google/gemma-3-12b-it list price $0.05/M input (the nearest hosted 12B Gemma; gemma-4-12b-it is not listed; the Winnow-12B / Jev-Omni precedent), read 2026-09-24, over the server's own usage.input_tokens; it is not a GPU bill.
Published source
https://github.com/blockbrain-ai/cygnet-recipe

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.