decider-chat-qwen3.6-27b: Qwen/Qwen3.6-27B read through the decider readout
This repository holds Qwen/Qwen3.6-27B (revision 6a9e13bd, weights unchanged,
55.6 GB in bf16) and the decider_config.json that reads it as a typed-decision model. There is no training. It is the
entry "Decider chat · Qwen3.6-27B" on the Decision Index.
- Input: a state and one or more typed questions (Choice, Noul yes/no, Score), each with an explicit option list.
- Output: a probability distribution over the options for every question, from one forward pass. There is no decoding.
- Readout: the model's chat template with thinking off, the option letter read at the answer slot, softmax over the option letters.
- Temperature: one temperature, T 1.943. The temperature was fitted by NLL on our own in-task rows, with the six tasks whose test splits are in the Decision Index suite excluded. An option-count rule gives this model almost nothing, so it keeps the single temperature.
Decision Index (edition v0.2.1, 2026-09-28, measured by the index on one RTX PRO 6000)
| entry | score (chance-corrected) | rank | ECE | latency, median |
|---|---|---|---|---|
| Decider chat · Qwen3.6-27B | 51.35 | #8 of 70 | 0.021 | 83.6 ms |
Jev 1.13.0 scores 57.91 on the same edition.
We replayed 300 stored rows of the index run through decider.serve with this repository's config. The package picked the
same answer as the stored run on 1,431 of 1,438 answers, with a median
probability difference of 0.0003.
Usage
Needs decider-ai>=1.5.0.
from decider.infer import Decider
d = Decider("Mapika/decider-chat-qwen3.6-27b")
d.system_one(state, {"refund": {"type": "noul", "instructions": "Is the refund allowed under the policy?",
"criteria": {"true": "allowed", "false": "not allowed"}}})
Serving: DECIDER_MODEL=Mapika/decider-chat-qwen3.6-27b uvicorn decider.serve:app (POST /v1/systemone). For large checkpoints,
decider.serve_vllm serves the same readout on vLLM (see the repository's docs/SERVING.md).
Limitations
The model is the stock base model. Its knowledge, reasoning and failure modes are the base model's. The readout adds a fixed answer slot and a fitted temperature, nothing else.
Licence
Weights: Qwen/Qwen3.6-27B, Apache-2.0 as published. Readout code: Apache-2.0 (github.com/Mapika/decider).