Save inference parity implementation and evaluation harness

src/server/request.rs

@@ -683,7 +683,7 @@ mod tests {
         );
         assert_eq!(
             request_reasoning(&default, ModelChoice::Qwen38FlashNext, "").unwrap(),
-            ReasoningMode::XHigh
+            ReasoningMode::Low
         );
 
         for (effort, mode) in [

src/server/response.rs

@@ -749,7 +749,7 @@ fn stream_response_with_keepalive(
                 ));
             }
             GenerationEvent::Activity(_) | GenerationEvent::Measured(_) => {}
-            GenerationEvent::Loading => {}
+            GenerationEvent::Loading | GenerationEvent::Loaded { .. } => {}
             GenerationEvent::Context {
                 tokens_per_second, ..
             } => {