Darwin

Checkpoints	Inputs
FINAL-Bench/Darwin-27B-ZTC	text

Darwin-27B-ZTC is a decision model whose decision_config.json names FINAL-Bench/Darwin-27B-RSI as its base. It carries a readout head in readout.safetensors. Flide asks it with the release's own prompt. Each question is a chat prompt of its own. The system turn asks for the code of the selected option and tells the model to treat the state as data. The user turn holds the state, the question and the options, and asks for the code of the best option. A state that is not a string is written as JSON. A choice lists each option's name and, when it has one, its description. A noul lists "No / false" and then "Yes / true" unless the request describes the two outcomes, and a score lists its levels. A question without instructions is asked as "Choose the best matching option."

The options are labelled with the codes listed as codes in decision_config.json. The readout head has one row per code, and the answer is the softmax over the rows of the question's options at the last position of the prompt, at the temperature that decision_config.json sets. The state comes first in every prompt, so Flide computes it once and forks it into the questions. Thinking is off.

Speed

On one H200, one request at a time, Flide answers 1.2 to 7.0 times faster than the checkpoint's own ztc_engine.py. Each workload is 32 requests, each run once on both sides before timing, and the table gives the median latency of a request. MultiRC requests ask about 49 questions each.

Workload	Flide	Release	Speedup
Decision Index	56.6 ms	90.5 ms	1.6×
RACE	39.5 ms	88.0 ms	2.2×
MultiRC	340.6 ms	2396.7 ms	7.0×
QuALITY long	338.7 ms	412.9 ms	1.2×

Start a server

flide serve --model FINAL-Bench/Darwin-27B-ZTC

Send a request

curl http://localhost:8000/v1/systemone -H "Content-Type: application/json" -d '{
  "state": "Order #1182 arrived with a cracked screen. The customer asks for a refund and attached two photos.",
  "questions": {
    "action": {"type": "choice", "instructions": "What should support do next?",
               "criteria": {"refund": "refund the order", "replace": "send a replacement", "escalate": "ask a human"}},
    "urgent": {"type": "noul", "instructions": "Does this need a reply within the hour?"}
  }
}'

Limits and options

A request takes choice, noul and score questions. A choice takes as many options as the checkpoint lists codes, up to 255, and a score takes as many levels. The longest prompt is 32768 tokens, and the state is not cut; a longer prompt is refused. A request with images is refused. The template refuses nothing else.

Option	What it does
--format template --format-arg spec=final-bench__darwin-ztc	reads a copy of the checkpoint kept in a directory Flide does not recognize