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

Bespoke Nimble 9B (Bespoke Labs)

Jev rebuild · by Bespoke Labs

JevBench v1.4.2.2 score

18.7

Rank #62 of 91 ranked systems.

44.6 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
46.3
calibration
56.4
speed
78.7
cost
33.4

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, Bespoke Nimble 9B vs Jev 1.13.0. Intelligence: 46.3 vs 53.1; Calibration: 56.4 vs 76.3; Speed: 78.7 vs 83.3; Cost: 33.4 vs 52.0.50100Intelligence46.3 · 53.1Calibration56.4 · 76.3Speed78.7 · 83.3Cost33.4 · 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, Bespoke Nimble 9B vs Jev 1.13.0. Easy: 100% vs 100%; Standard: 95% vs 99%; Judge: 89% vs 95%; Hard: 65% vs 74%; Sealed: 29% vs 37%.50100Easy100% · 100%Standard95% · 99%Judge89% · 95%Hard65% · 74%Sealed29% · 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), Bespoke Nimble 9B vs Jev 1.13.0. Ambiguous / abstain: 33% vs 43%; Judge: 50% vs 54%; Long policy: 41% vs 44%; Multi-hop: 53% vs 64%; Probability: 48% vs 63%; Temporal / numeric: 23% vs 28%; Trade-off: 37% vs 55%; Routing: 90% vs 100%; Trap / adversarial: 75% vs 83%; Paraphrase: 50% vs 64%; Safety judge: 31% vs 38%.50100Ambiguous /abstain33% · 43%Judge50% · 54%Long policy41% · 44%Multi-hop53% · 64%Probability48% · 63%Temporal /numeric23% · 28%Trade-off37% · 55%Routing90% · 100%Trap /adversarial75% · 83%Paraphrase50% · 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, Bespoke Nimble 9B vs Jev 1.13.0. Ambiguous / abstain: 27% vs 30%; Judge: 29% vs 34%; Long policy: 23% vs 28%; Multi-hop: 29% vs 45%; Paraphrase: 50% vs 64%; Probability: 39% vs 50%; Safety judge: 31% vs 38%; Temporal / numeric: 20% vs 29%; Trade-off: 27% vs 38%; Trap / adversarial: 50% vs 42%.50100Ambiguous /abstain27% · 30%Judge29% · 34%Long policy23% · 28%Multi-hop29% · 45%Paraphrase50% · 64%Probability39% · 50%Safety judge31% · 38%Temporal /numeric20% · 29%Trade-off27% · 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 (weights); repository without a licence file as of 19 Sep
Cost evidence
estimated; the board’s row disclosure contains the published basis.
Endpoint condition
our RunPod GPU (A40 48 GB, Canada), reached over the internet from Germany
Note on this row
Re-run in v1.2.8 at Bespoke Labs' request after they raised the serving prompt limit from 2,048 to 8,192 tokens (bespokelabsai/nimble PR #4). Same recipe as the v1.1.3 run — the published LoRA merged into Qwen3.5-9B with the author's PEFT safe-merge, served with SGLang and the author's Jev-compatible API — now from current nimble main; the adapter weights are unchanged. Hard-tier accuracy rose from 43.6 % to 65.5 %, yet the score fell: the long hard items that used to fail at once are now answered and priced (so Cost fell), and this pod was in Canada while the v1.1.3 run's was in Sweden, so part of the lower Speed is network distance from our server in Germany. This complete run replaces the earlier row; its old score is kept in the artifact under superseded_rows.
Published source
https://github.com/bespokelabsai/nimble

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.