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.
Metask Lab's calibrated decision model in 16 languages. Qwen3.5-4B with a merged LoRA trained with the Nimble candidate-logit objective; one forward pass and a softmax over at most 26 answer-letter logits, with one temperature per question type.
Decides
choice, score, noul, classify, route
choice, score, noul, classify, route
Architecture
anyjev
metask-jev
Fine-tuned from
qwen/qwen3-8b
qwen/qwen3.5-4b
License
apache-2.0
apache-2.0
Availability
Open weights
Open weights
Hosted by
—
—
Input price
—
—
Decision accuracy
77.1%
80.1%
Calibration error
0.034
—
Valid action rate
—
—
Median latency
—
62.8 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 metask-jev?
anyjev is from Nokia and metask-jev from Metask Lab. Both have open weights you can download and run. Both answer choice, score, noul, classify and route questions. metask-jev is the smaller model, at 4.5B parameters to 8.0B.
Which is more accurate, anyjev or metask-jev?
They report on different suites — anyjev 77.1% on LocalLLaMA/typed-decisions (2,000 held-out decisions), L2 with 300 labels per question, metask-jev 80.1% on JevBench v1.2 public set (231 items) at 4,096 tokens, maker's run — so the numbers do not rank them. Test both on your own labelled examples.
Which is cheaper, anyjev or metask-jev?
anyjev: Free (open weights). metask-jev: Free (open weights). Open weights cost nothing per call beyond your own hardware.
Can I run anyjev or metask-jev locally?
Yes, both: systemone pull nokia/anyjev and systemone pull metask-lab/metask-jev download the weights.
Evaluation suite
LocalLLaMA/typed-decisions (2,000 held-out decisions), L2 with 300 labels per question
JevBench v1.2 public set (231 items) at 4,096 tokens, maker's run
