diff --git a/codex-rs/core/src/client.rs b/codex-rs/core/src/client.rs
index 701aa3e4c2..3c7a645a18 100644
--- a/codex-rs/core/src/client.rs
+++ b/codex-rs/core/src/client.rs
@@ -112,6 +112,8 @@ impl ModelClient {
     /// For Chat providers, the underlying stream is optionally aggregated
     /// based on the `show_raw_agent_reasoning` flag in the config.
     pub async fn stream(&self, prompt: &PromptBuilder) -> Result<ResponseStream> {
+        self.validate_prompt_wire_api(prompt)?;
+
         match self.provider.wire_api {
             WireApi::Responses => self.stream_responses_api(prompt).await,
             WireApi::Chat => {
@@ -132,6 +134,17 @@ impl ModelClient {
         }
     }
 
+    fn validate_prompt_wire_api(&self, prompt: &PromptBuilder) -> Result<()> {
+        if prompt.wire_api != self.provider.wire_api {
+            return Err(CodexErr::UnsupportedOperation(format!(
+                "prompt configured for {:?} wire API but provider {} expects {:?}",
+                prompt.wire_api, self.provider.name, self.provider.wire_api
+            )));
+        }
+
+        Ok(())
+    }
+
     /// Streams a turn via the OpenAI Chat Completions API.
     ///
     /// This path is only used when the provider is configured with
diff --git a/codex-rs/core/src/client_common.rs b/codex-rs/core/src/client_common.rs
index 2afac467de..4d28deea29 100644
--- a/codex-rs/core/src/client_common.rs
+++ b/codex-rs/core/src/client_common.rs
@@ -5,6 +5,8 @@ use crate::model_family::ModelFamily;
 use crate::model_provider_info::WireApi;
 use crate::tools::spec::create_tools_json_for_chat_completions_api;
 use crate::tools::spec::create_tools_json_for_responses_api;
+use codex_api::CompactionInput;
+use codex_api::Prompt;
 pub use codex_api::common::ResponseEvent;
 use codex_apply_patch::APPLY_PATCH_TOOL_INSTRUCTIONS;
 use codex_protocol::models::ResponseItem;
@@ -18,7 +20,6 @@ use std::pin::Pin;
 use std::task::Context;
 use std::task::Poll;
 use tokio::sync::mpsc;
-use codex_api::{CompactionInput, Prompt};
 
 /// Review thread system prompt. Edit `core/src/review_prompt.md` to customize.
 pub const REVIEW_PROMPT: &str = include_str!("../review_prompt.md");
diff --git a/codex-rs/core/tests/chat_completions_payload.rs b/codex-rs/core/tests/chat_completions_payload.rs
index c66bbdc78f..24c7585c10 100644
--- a/codex-rs/core/tests/chat_completions_payload.rs
+++ b/codex-rs/core/tests/chat_completions_payload.rs
@@ -10,6 +10,7 @@ use codex_core::ModelProviderInfo;
 use codex_core::PromptBuilder;
 use codex_core::ResponseItem;
 use codex_core::WireApi;
+use codex_core::error::CodexErr;
 use codex_otel::otel_event_manager::OtelEventManager;
 use codex_protocol::ConversationId;
 use codex_protocol::models::ReasoningItemContent;
@@ -41,21 +42,7 @@ async fn run_request(input: Vec<ResponseItem>) -> Value {
         .mount(&server)
         .await;
 
-    let provider = ModelProviderInfo {
-        name: "mock".into(),
-        base_url: Some(format!("{}/v1", server.uri())),
-        env_key: None,
-        env_key_instructions: None,
-        experimental_bearer_token: None,
-        wire_api: WireApi::Chat,
-        query_params: None,
-        http_headers: None,
-        env_http_headers: None,
-        request_max_retries: Some(0),
-        stream_max_retries: Some(0),
-        stream_idle_timeout_ms: Some(5_000),
-        requires_openai_auth: false,
-    };
+    let provider = chat_provider(format!("{}/v1", server.uri()));
 
     let codex_home = match TempDir::new() {
         Ok(dir) => dir,
@@ -184,6 +171,24 @@ fn first_assistant(messages: &[Value]) -> &Value {
     }
 }
 
+fn chat_provider(base_url: String) -> ModelProviderInfo {
+    ModelProviderInfo {
+        name: "mock".into(),
+        base_url: Some(base_url),
+        env_key: None,
+        env_key_instructions: None,
+        experimental_bearer_token: None,
+        wire_api: WireApi::Chat,
+        query_params: None,
+        http_headers: None,
+        env_http_headers: None,
+        request_max_retries: Some(0),
+        stream_max_retries: Some(0),
+        stream_idle_timeout_ms: Some(5_000),
+        requires_openai_auth: false,
+    }
+}
+
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
 async fn omits_reasoning_when_none_present() {
     skip_if_no_network!();
@@ -316,3 +321,52 @@ async fn suppresses_duplicate_assistant_messages() {
         Value::String("dup".into())
     );
 }
+
+#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+async fn errors_on_mismatched_prompt_wire_api() {
+    let provider = chat_provider("https://example.com/v1".into());
+
+    let codex_home = match TempDir::new() {
+        Ok(dir) => dir,
+        Err(e) => panic!("failed to create TempDir: {e}"),
+    };
+    let mut config = load_default_config_for_test(&codex_home);
+    config.model_provider_id = provider.name.clone();
+    config.model_provider = provider.clone();
+    config.show_raw_agent_reasoning = true;
+    let effort = config.model_reasoning_effort;
+    let summary = config.model_reasoning_summary;
+    let config = Arc::new(config);
+
+    let conversation_id = ConversationId::new();
+
+    let otel_event_manager = OtelEventManager::new(
+        conversation_id,
+        config.model.as_str(),
+        config.model_family.slug.as_str(),
+        None,
+        Some("test@test.com".to_string()),
+        Some(AuthMode::ChatGPT),
+        false,
+        "test".to_string(),
+    );
+
+    let client = ModelClient::new(
+        Arc::clone(&config),
+        None,
+        otel_event_manager,
+        provider,
+        effort,
+        summary,
+        conversation_id,
+        codex_protocol::protocol::SessionSource::Exec,
+    );
+
+    let prompt = PromptBuilder::new().with_input(vec![user_message("u1")]);
+
+    let err = match client.stream(&prompt).await {
+        Ok(_) => panic!("wire API mismatch should error before sending request"),
+        Err(e) => e,
+    };
+    assert!(matches!(err, CodexErr::UnsupportedOperation(_)));
+}