diff --git a/rust/Cargo.toml b/rust/Cargo.toml
index 344f10b3f28..435350c0711 100644
--- a/rust/Cargo.toml
+++ b/rust/Cargo.toml
@@ -144,6 +144,13 @@ too_many_arguments = "allow"
 [profile.dev]
 panic = "abort"
 
+# Speed up cold tokenizer construction in tests.
+[profile.dev.package]
+fastokens = { opt-level = 3 }
+regex-automata = { opt-level = 3 }
+serde_json = { opt-level = 3 }
+tokenizers = { opt-level = 3 }
+
 [profile.release]
 lto = "thin"
 panic = "abort"
diff --git a/rust/src/chat/tests/roundtrip.rs b/rust/src/chat/tests/roundtrip.rs
index f2ad815af44..e6044614aa9 100644
--- a/rust/src/chat/tests/roundtrip.rs
+++ b/rust/src/chat/tests/roundtrip.rs
@@ -214,14 +214,17 @@ macro_rules! roundtrip_tests {
     ($($case:ident => [$($(#[$fixture_attr:meta])* $fixture:ident),* $(,)?]),+ $(,)?) => {
         paste::paste! {
             $(
-                $(
-                    #[tokio::test]
-                    $(#[$fixture_attr])*
-                    #[file_serial([<hf_ $case>])]
-                    async fn [<roundtrip_ $case _ $fixture>]() -> Result<()> {
-                        [<run_roundtrip_ $fixture>](RoundtripCase::$case()).await
-                    }
-                )*
+                #[tokio::test]
+                #[file_serial([<hf_ $case>])]
+                async fn [<roundtrip_ $case>]() -> Result<()> {
+                    let case = RoundtripCase::$case();
+                    let backends = load_roundtrip_backends(&case).await?;
+                    $(
+                        $(#[$fixture_attr])*
+                        [<run_roundtrip_ $fixture>](&case, &backends).await?;
+                    )*
+                    Ok(())
+                }
             )+
         }
     };
@@ -241,18 +244,21 @@ roundtrip_tests! {
 }
 
 /// Run the fixed reasoning+content fixture for one model/parser case.
-async fn run_roundtrip_reasoning_and_content(case: RoundtripCase) -> Result<()> {
+async fn run_roundtrip_reasoning_and_content(
+    case: &RoundtripCase,
+    backends: &vllm_chat::LoadedModelBackends,
+) -> Result<()> {
     for thinking in case.thinking_behavior.fixtures() {
-        run_roundtrip_reasoning_and_content_inner(case.clone(), thinking).await?;
+        run_roundtrip_reasoning_and_content_inner(case, backends, thinking).await?;
     }
     Ok(())
 }
 
 async fn run_roundtrip_reasoning_and_content_inner(
-    case: RoundtripCase,
+    case: &RoundtripCase,
+    backends: &vllm_chat::LoadedModelBackends,
     thinking: Option<bool>,
 ) -> Result<()> {
-    let backends = load_roundtrip_backends(&case).await?;
     let request = roundtrip_request(
         "roundtrip-reasoning-content",
         vec![ChatMessage::text(ChatRole::User, "What is 2 + 2?")],
@@ -275,7 +281,7 @@ async fn run_roundtrip_reasoning_and_content_inner(
         });
         AssistantMessage { content }
     };
-    let result = run_roundtrip(&case, &backends, &request, assistant).await?;
+    let result = run_roundtrip(case, backends, &request, assistant).await?;
 
     assert_eq!(
         result.parsed_message.reasoning().as_deref().map(str::trim),
@@ -293,8 +299,10 @@ async fn run_roundtrip_reasoning_and_content_inner(
 }
 
 /// Run the fixed reasoning+multiple-tools fixture for one model/parser case.
-async fn run_roundtrip_tool_call_mix(case: RoundtripCase) -> Result<()> {
-    let backends = load_roundtrip_backends(&case).await?;
+async fn run_roundtrip_tool_call_mix(
+    case: &RoundtripCase,
+    backends: &vllm_chat::LoadedModelBackends,
+) -> Result<()> {
     let request = roundtrip_request(
         "roundtrip-reasoning-tools",
         vec![ChatMessage::text(
@@ -308,8 +316,8 @@ async fn run_roundtrip_tool_call_mix(case: RoundtripCase) -> Result<()> {
     let expected_text = "I will call the tools.";
 
     let result = run_roundtrip(
-        &case,
-        &backends,
+        case,
+        backends,
         &request,
         AssistantMessage {
             content: vec![
@@ -353,12 +361,12 @@ async fn run_roundtrip_tool_call_mix(case: RoundtripCase) -> Result<()> {
     assert_eq!(tool_calls[0].name, "get_weather");
     assert_eq!(
         tool_calls[0].arguments,
-        expected_arguments(&case, r#"{"location": "Shanghai"}"#)?,
+        expected_arguments(case, r#"{"location": "Shanghai"}"#)?,
     );
     assert_eq!(tool_calls[1].name, "add");
     assert_eq!(
         tool_calls[1].arguments,
-        expected_arguments(&case, r#"{"y": 1.0, "x": 2, "items": ["left", "right"]}"#)?,
+        expected_arguments(case, r#"{"y": 1.0, "x": 2, "items": ["left", "right"]}"#)?,
     );
 
     assert_eq!(