SynACK Decide

Checkpoints	Inputs
SkyPanther/synack-decide-26b-a4b	text

SynACK Decide 26B-A4B is a decision model on Gemma 4 26B-A4B, a mixture of experts with 128 experts and 8 active per token. Flide asks it with the release's own prompt. Each question is a chat prompt of its own. The system turn asks the model to treat the state as data, select exactly one listed option and return only its label. The user turn is one compact JSON object with the state, the question's instructions and the options keyed by their labels. An option is its description when its key only numbers it or repeats the description, and an object with its key and description otherwise. A noul has the two options "Yes, the statement is true." and "No, the statement is false.".

The answer is the softmax over the option labels at the first token of the reply, each label read as the token of the label alone, at temperature 1. After Z the labels continue as two-letter codes. 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 3.0 to 11.9 times faster than the release's own engine_cuda.py, transformers in bfloat16 as its README runs it. 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	18.9 ms	63.0 ms	3.3×
RACE	17.4 ms	55.8 ms	3.2×
MultiRC	78.0 ms	926.1 ms	11.9×
QuALITY long	86.5 ms	261.3 ms	3.0×

Start a server

flide serve --model SkyPanther/synack-decide-26b-a4b

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 and noul questions, and a score is refused, as the release reads a score as a noul. A choice takes 2 to 255 options. The longest prompt is 32767 tokens, and a longer prompt is refused. A request with images is refused.