diff --git a/crates/switchyard-translation/src/codecs/openai_chat/buffered.rs b/crates/switchyard-translation/src/codecs/openai_chat/buffered.rs
index 0dc0641625..4e00f5d0a6 100644
--- a/crates/switchyard-translation/src/codecs/openai_chat/buffered.rs
+++ b/crates/switchyard-translation/src/codecs/openai_chat/buffered.rs
@@ -1292,7 +1292,7 @@ pub(crate) fn encode_openai_usage(usage: &Usage) -> Value {
     if usage.cached_input_tokens().is_some() || usage.cache_creation_input_tokens().is_some() {
         value["prompt_tokens_details"] = json!({
             "cached_tokens": usage.cached_input_tokens().unwrap_or(0),
-            "cache_creation_tokens": usage.cache_creation_input_tokens().unwrap_or(0),
+            "cache_write_tokens": usage.cache_creation_input_tokens().unwrap_or(0),
         });
     }
     if let Some(reasoning_tokens) = usage.reasoning_tokens {
diff --git a/crates/switchyard-translation/tests/response_translation.rs b/crates/switchyard-translation/tests/response_translation.rs
index cef7155dd7..7d3ce5df57 100644
--- a/crates/switchyard-translation/tests/response_translation.rs
+++ b/crates/switchyard-translation/tests/response_translation.rs
@@ -377,7 +377,7 @@ fn responses_reasoning_usage_translates_to_openai_chat_usage_details() -> TestRe
         json!({"reasoning_tokens": 3})
     );
     assert_eq!(
-        output["usage"]["prompt_tokens_details"]["cache_creation_tokens"],
+        output["usage"]["prompt_tokens_details"]["cache_write_tokens"],
         2
     );
     let decoded = engine