diff --git a/atoma-service/src/handlers/chat_completions.rs b/atoma-service/src/handlers/chat_completions.rs
index 48d58162..a97ed236 100644
--- a/atoma-service/src/handlers/chat_completions.rs
+++ b/atoma-service/src/handlers/chat_completions.rs
@@ -34,7 +34,7 @@ use crate::{
     middleware::RequestMetadata,
 };
 
-use super::handle_confidential_compute_encryption_response;
+use super::{handle_confidential_compute_encryption_response, handle_status_code_error};
 
 /// The path for confidential chat completions requests
 pub const CONFIDENTIAL_CHAT_COMPLETIONS_PATH: &str = "/v1/confidential/chat/completions";
@@ -683,10 +683,11 @@ async fn handle_streaming_response(
         })?;
 
     if !response.status().is_success() {
-        return Err(AtomaServiceError::InternalError {
-            message: "Inference service returned error".to_string(),
-            endpoint,
-        });
+        let error = response
+            .status()
+            .canonical_reason()
+            .unwrap_or("Unknown error");
+        handle_status_code_error(response.status(), &endpoint, error)?;
     }
 
     let stream = response.bytes_stream();
@@ -1066,6 +1067,8 @@ pub mod utils {
     use atoma_utils::constants::PAYLOAD_HASH_SIZE;
     use prometheus::HistogramTimer;
 
+    use crate::handlers::handle_status_code_error;
+
     use super::{
         handle_confidential_compute_encryption_response, info, instrument,
         sign_response_and_update_stack_hash, update_stack_num_compute_units, AppState,
@@ -1271,13 +1274,11 @@ pub mod utils {
         })?;
 
         if !response.status().is_success() {
-            return Err(AtomaServiceError::InternalError {
-                message: format!(
-                    "Inference service returned non-success status code: {}",
-                    response.status()
-                ),
-                endpoint: endpoint.to_string(),
-            });
+            let error = response
+                .status()
+                .canonical_reason()
+                .unwrap_or("Unknown error");
+            handle_status_code_error(response.status(), endpoint, error)?;
         }
 
         response.json::<Value>().await.map_err(|e| {
diff --git a/atoma-service/src/handlers/embeddings.rs b/atoma-service/src/handlers/embeddings.rs
index 4c9b28a2..7a016eb6 100644
--- a/atoma-service/src/handlers/embeddings.rs
+++ b/atoma-service/src/handlers/embeddings.rs
@@ -19,6 +19,8 @@ use serde_json::Value;
 use tracing::{info, instrument};
 use utoipa::OpenApi;
 
+use super::handle_status_code_error;
+
 /// The path for confidential embeddings requests
 pub const CONFIDENTIAL_EMBEDDINGS_PATH: &str = "/v1/confidential/embeddings";
 
@@ -321,13 +323,11 @@ async fn handle_embeddings_response(
         })?;
 
     if !response.status().is_success() {
-        return Err(AtomaServiceError::InternalError {
-            message: format!(
-                "Inference service returned non-success status code: {}",
-                response.status()
-            ),
-            endpoint: endpoint.to_string(),
-        });
+        let error = response
+            .status()
+            .canonical_reason()
+            .unwrap_or("Unknown error");
+        handle_status_code_error(response.status(), endpoint, error)?;
     }
 
     let mut response_body =
diff --git a/atoma-service/src/handlers/image_generations.rs b/atoma-service/src/handlers/image_generations.rs
index d19bedc5..ee7085c0 100644
--- a/atoma-service/src/handlers/image_generations.rs
+++ b/atoma-service/src/handlers/image_generations.rs
@@ -18,7 +18,10 @@ use serde_json::Value;
 use tracing::{info, instrument};
 use utoipa::OpenApi;
 
-use super::{handle_confidential_compute_encryption_response, sign_response_and_update_stack_hash};
+use super::{
+    handle_confidential_compute_encryption_response, handle_status_code_error,
+    sign_response_and_update_stack_hash,
+};
 
 /// The path for confidential image generations requests
 pub const CONFIDENTIAL_IMAGE_GENERATIONS_PATH: &str = "/v1/confidential/images/generations";
@@ -307,13 +310,11 @@ async fn handle_image_generations_response(
         })?;
 
     if !response.status().is_success() {
-        return Err(AtomaServiceError::InternalError {
-            message: format!(
-                "Inference service returned non-success status code: {}",
-                response.status()
-            ),
-            endpoint: endpoint.to_string(),
-        });
+        let error = response
+            .status()
+            .canonical_reason()
+            .unwrap_or("Unknown error");
+        handle_status_code_error(response.status(), endpoint, error)?;
     }
 
     let mut response_body =
diff --git a/atoma-service/src/handlers/mod.rs b/atoma-service/src/handlers/mod.rs
index 4ba00449..4f4e0676 100644
--- a/atoma-service/src/handlers/mod.rs
+++ b/atoma-service/src/handlers/mod.rs
@@ -9,6 +9,7 @@ use atoma_confidential::types::{
 use atoma_utils::hashing::blake2b_hash;
 use base64::{engine::general_purpose::STANDARD, Engine};
 use flume::Sender;
+use hyper::StatusCode;
 use image_generations::CONFIDENTIAL_IMAGE_GENERATIONS_PATH;
 use serde_json::{json, Value};
 use tracing::{info, instrument};
@@ -305,3 +306,42 @@ pub fn update_stack_num_compute_units(
             endpoint: endpoint.to_string(),
         })
 }
+
+/// Handles the status code returned by the inference service.
+///
+/// This function maps the status code to an appropriate error variant.
+///
+/// # Arguments
+///
+/// * `status_code` - The status code returned by the inference service
+/// * `endpoint` - The API endpoint path where the request was received
+/// * `error` - The error message returned by the inference service
+///
+/// # Returns
+///
+/// Returns an `AtomaServiceError` variant based on the status code.
+#[instrument(level = "info", skip_all, fields(endpoint))]
+pub fn handle_status_code_error(
+    status_code: StatusCode,
+    endpoint: &str,
+    error: &str,
+) -> Result<(), AtomaServiceError> {
+    match status_code {
+        StatusCode::UNAUTHORIZED => Err(AtomaServiceError::AuthError {
+            auth_error: format!("Unauthorized response from inference service: {error}"),
+            endpoint: endpoint.to_string(),
+        }),
+        StatusCode::INTERNAL_SERVER_ERROR => Err(AtomaServiceError::InternalError {
+            message: format!("Inference service returned internal server error: {error}"),
+            endpoint: endpoint.to_string(),
+        }),
+        StatusCode::BAD_REQUEST => Err(AtomaServiceError::InvalidBody {
+            message: format!("Inference service returned bad request error: {error}"),
+            endpoint: endpoint.to_string(),
+        }),
+        _ => Err(AtomaServiceError::InternalError {
+            message: format!("Inference service returned non-success error: {error}"),
+            endpoint: endpoint.to_string(),
+        }),
+    }
+}