Clarify model-only release and serving guidance

Signed-off-by: Xunzhuo <[email protected]>

README.md

@@ -94,7 +94,7 @@ The published model's complete state, question, and candidates have a 16,384-tok
 
 ![Decision decoder architecture](assets/architecture.png)
 
-A causal Qwen3.5 text backbone combines gated linear and full attention. A shared candidate head reads candidate endpoints and the final query vector. Each question uses one forward pass; questions run independently in batches of eight.
+A causal Qwen3.5 text backbone combines gated linear and full attention. A shared candidate head reads candidate endpoints and the final query vector. The serving runtime schedules questions according to available hardware and request load.
 
 [Candidate head](assets/readout.png) · [Vector architecture](assets/architecture.svg)