A training-free layer that turns an open LLM into a decision model. It reads typed choice, yes/no and score answers from one prefill, removes option-order bias with no labels and, from a few hundred labels, calibrates or fits a closed-form head. Weights stay untouched.
An open Jev-style LoRA on Qwen3.5-9B from Bespoke Labs, trained on 2,676 contrastively curated examples to score the allowed answer tokens directly for enums, booleans and rubric levels. Recipe, data and a public benchmark suite are released with it.
Decides
choice, score, noul, classify, route
choice, noul, score, classify, route
Architecture
anyjev
nimble
Fine-tuned from
qwen/qwen3-8b
qwen/qwen3.5-9b
License
apache-2.0
apache-2.0
Availability
Open weights
Open weights + hosted API
Hosted by
—
Bespoke Labs
Input price
—
—
Decision accuracy
77.1%
90.1%
Calibration error
0.034
0.054
Valid action rate
—
—
Median latency
—
106 ms
p95 latency
—
—
Figures are from each model’s manifest; accuracy and latency are what the publishers report, on their own suites and hardware. Add a third model.
Questions
What is the difference between anyjev and bespoke-nimble-9b?
anyjev is from Nokia Applied Research and bespoke-nimble-9b from Bespoke Labs. anyjev has open weights you can download and run; bespoke-nimble-9b has open weights and a hosted API. Both answer choice, score, noul, classify and route questions. anyjev is the smaller model, at 8.0B parameters to 9.0B.
Which is more accurate, anyjev or bespoke-nimble-9b?
They report on different suites — anyjev 77.1% on LocalLLaMA/typed-decisions (2,000 held-out decisions), L2 with 300 labels per question, bespoke-nimble-9b 90.1% on Bespoke held-out set (324 examples) — so the numbers do not rank them. Test both on your own labelled examples.
Which is cheaper, anyjev or bespoke-nimble-9b?
anyjev: Free (open weights). bespoke-nimble-9b: Hosted, price not published, or free to self-host. Open weights cost nothing per call beyond your own hardware.
Can I run anyjev or bespoke-nimble-9b locally?
Yes, both: systemone pull nokia-applied-research/anyjev and systemone pull bespoke-labs/bespoke-nimble-9b download the weights.
Evaluation suite
LocalLLaMA/typed-decisions (2,000 held-out decisions), L2 with 300 labels per question
