A 0.4B decision model made from the first 20 layers of Qwen3-0.6B plus a small attention head that compares options. It answers choice, noul and score questions in one pass, and choice order cannot change its answer by construction. Non-commercial licence.
TokenRhythm's prefill-only decision model for agent workflows. Built on its NeoHorse-1-4B (a Qwen3.5-4B derivative), it predicts Choice, Noul and Score answers over application-defined options without generating text, with optional single-image input.
Decides
choice, score, noul
choice, score, noul, classify, route
Architecture
bev-decider
neohorse
Fine-tuned from
qwen/qwen3-0.6b
tokenrhythm/neohorse-1-4b
License
cc-by-nc-4.0
apache-2.0
Availability
Open weights
Open weights
Hosted by
—
—
Input price
—
—
Decision accuracy
74.7%
75.3%
Calibration error
—
—
Valid action rate
—
—
Median latency
—
—
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 bev-decider and neohorse-jev?
bev-decider is from Avishek Biswas and neohorse-jev from TokenRhythm. Both have open weights you can download and run. Both answer choice, score and noul questions. Only neohorse-jev answers classify and route. bev-decider is the smaller model, at 478M parameters to 4.0B. bev-decider is licensed cc-by-nc-4.0; neohorse-jev, apache-2.0.
Which is more accurate, bev-decider or neohorse-jev?
They report on different suites — bev-decider 74.7% on avbiswas/bev-decision test split (5,000 held-out questions over 2,617 states), the maker's own, neohorse-jev 75.3% on JevBench public set (231 items), vLLM, maker's run — so the numbers do not rank them. Test both on your own labelled examples.
Which is cheaper, bev-decider or neohorse-jev?
bev-decider: Free (open weights). neohorse-jev: Free (open weights). Open weights cost nothing per call beyond your own hardware.
Can I run bev-decider or neohorse-jev locally?
Yes, both: systemone pull avishek-biswas/bev-decider and systemone pull tokenrhythm/neohorse-jev download the weights.
—
Evaluation suite
avbiswas/bev-decision test split (5,000 held-out questions over 2,617 states), the maker's own
JevBench public set (231 items), vLLM, maker's run
