diff --git a/src-tauri/Cargo.lock b/src-tauri/Cargo.lock
index 60bd57a94..13b977ea2 100644
--- a/src-tauri/Cargo.lock
+++ b/src-tauri/Cargo.lock
@@ -825,6 +825,7 @@ dependencies = [
  "windows-sys 0.61.2",
  "winreg 0.52.0",
  "zip 2.4.2",
+ "zstd",
 ]
 
 [[package]]
diff --git a/src-tauri/Cargo.toml b/src-tauri/Cargo.toml
index 9f7841511..9fc61c123 100644
--- a/src-tauri/Cargo.toml
+++ b/src-tauri/Cargo.toml
@@ -43,6 +43,7 @@ reqwest = { version = "0.12", features = ["rustls-tls", "json", "stream", "socks
 arboard = "3.6"
 flate2 = "1"
 brotli = "7"
+zstd = "0.13"
 tokio = { version = "1", features = ["macros", "rt-multi-thread", "time", "sync"] }
 futures = "0.3"
 async-stream = "0.3"
diff --git a/src-tauri/src/proxy/content_encoding.rs b/src-tauri/src/proxy/content_encoding.rs
new file mode 100644
index 000000000..970ffd289
--- /dev/null
+++ b/src-tauri/src/proxy/content_encoding.rs
@@ -0,0 +1,234 @@
+//! HTTP content-encoding 工具。
+//!
+//! reqwest 的自动解压已禁用（为了透传 accept-encoding），需要手动解压。
+//! 请求侧（如 Codex Desktop 在登录态发压缩请求体）与响应侧（上游压缩响应体）
+//! 共用同一套解压逻辑。
+
+use axum::http::header::HeaderMap;
+use std::io::Read;
+
+/// 把 content-encoding 值拆成有序 coding 列表（去掉 identity 与空值）。
+///
+/// HTTP 允许堆叠编码（如 `gzip, zstd`），各 coding 以逗号分隔；亦允许重复
+/// content-encoding 头，语义等同逗号拼接（见 [`get_content_encoding`]）。
+fn split_codings(content_encoding: &str) -> Vec<&str> {
+    content_encoding
+        .split(',')
+        .map(str::trim)
+        .filter(|c| !c.is_empty() && *c != "identity")
+        .collect()
+}
+
+/// 单个 coding 是否可被解压。
+fn is_single_supported(coding: &str) -> bool {
+    matches!(
+        coding,
+        "gzip" | "x-gzip" | "deflate" | "br" | "zstd" | "zst"
+    )
+}
+
+/// 解压单个 content-coding。未知编码返回 `Ok(None)`。
+fn decompress_single(coding: &str, body: &[u8]) -> Result<Option<Vec<u8>>, std::io::Error> {
+    match coding {
+        "gzip" | "x-gzip" => {
+            let mut decoder = flate2::read::GzDecoder::new(body);
+            let mut decompressed = Vec::new();
+            decoder.read_to_end(&mut decompressed)?;
+            Ok(Some(decompressed))
+        }
+        "deflate" => {
+            // RFC 9110: deflate 指 zlib 包裹格式；但部分上游 / 客户端发 raw deflate 流。
+            // 先按规范尝试 zlib，失败再回退 raw —— 否则合规来源必然解压失败，
+            // 原始压缩字节会被 fail-open 透传给 JSON 解析（#2234 形态 C 之一）。
+            let mut decompressed = Vec::new();
+            let mut zlib = flate2::read::ZlibDecoder::new(body);
+            match zlib.read_to_end(&mut decompressed) {
+                Ok(_) => Ok(Some(decompressed)),
+                Err(zlib_err) => {
+                    log::debug!("deflate 按 zlib 解压失败（{zlib_err}），回退 raw deflate");
+                    let mut decompressed = Vec::new();
+                    let mut raw = flate2::read::DeflateDecoder::new(body);
+                    raw.read_to_end(&mut decompressed)?;
+                    Ok(Some(decompressed))
+                }
+            }
+        }
+        "br" => {
+            let mut decompressed = Vec::new();
+            brotli::BrotliDecompress(&mut std::io::Cursor::new(body), &mut decompressed)?;
+            Ok(Some(decompressed))
+        }
+        "zstd" | "zst" => {
+            // Codex 登录态对请求体启用 zstd（Compression::Zstd）；上游也可能 zstd 压缩响应。
+            let decompressed = zstd::stream::decode_all(std::io::Cursor::new(body))?;
+            Ok(Some(decompressed))
+        }
+        _ => Ok(None),
+    }
+}
+
+/// 根据 content-encoding 解压 body 字节，支持堆叠编码（如 `gzip, zstd`）。
+///
+/// RFC 9110 §8.4：codings 按**应用顺序**列出，故解压须**反向**（最后应用的先解）。
+/// 返回 `Ok(None)` 表示存在不受支持的编码、原样透传——此时调用方必须保留
+/// content-encoding 头，否则下游（诊断 / 客户端）会把压缩字节误当明文。
+pub(crate) fn decompress_body(
+    content_encoding: &str,
+    body: &[u8],
+) -> Result<Option<Vec<u8>>, std::io::Error> {
+    let codings = split_codings(content_encoding);
+    if codings.is_empty() {
+        return Ok(None);
+    }
+    // 任一 coding 不支持就整体放弃解压、保头透传，避免半解码的脏数据。
+    if !codings.iter().all(|c| is_single_supported(c)) {
+        log::warn!("不支持的 content-encoding: {content_encoding}，跳过解压");
+        return Ok(None);
+    }
+
+    // 反向解码：列表末尾是最后应用的编码，须最先解。
+    let mut data: Option<Vec<u8>> = None;
+    for coding in codings.iter().rev() {
+        let input = data.as_deref().unwrap_or(body);
+        match decompress_single(coding, input)? {
+            Some(decompressed) => data = Some(decompressed),
+            // 上面 is_single_supported 已校验，理论不会发生；防御性兜底。
+            None => return Ok(None),
+        }
+    }
+    Ok(data)
+}
+
+/// 该 content-encoding（含堆叠，如 `gzip, zstd`）是否全部可被解压。
+///
+/// 请求侧用它做闸门：无法解压的压缩体不能透传给 JSON 解析，需直接拒绝。
+pub(crate) fn is_supported_content_encoding(content_encoding: &str) -> bool {
+    let codings = split_codings(content_encoding);
+    !codings.is_empty() && codings.iter().all(|c| is_single_supported(c))
+}
+
+/// 从 header 提取 content-encoding（合并重复头，忽略 identity 与空值）。
+///
+/// HTTP 允许重复 content-encoding 头，语义等同逗号拼接，故用 `get_all` 合并；
+/// 返回值可能含多个逗号分隔的 coding，交由 [`decompress_body`] 反向解码。
+pub(crate) fn get_content_encoding(headers: &HeaderMap) -> Option<String> {
+    let combined = headers
+        .get_all("content-encoding")
+        .iter()
+        .filter_map(|v| v.to_str().ok())
+        .map(str::trim)
+        .filter(|s| !s.is_empty())
+        .collect::<Vec<_>>()
+        .join(", ")
+        .to_lowercase();
+    if split_codings(&combined).is_empty() {
+        return None;
+    }
+    Some(combined)
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use axum::http::HeaderValue;
+
+    #[test]
+    fn decompress_body_deflate_handles_zlib_wrapped_per_rfc9110() {
+        // RFC 9110 规范的 deflate = zlib 包裹格式（合规来源发的就是这个）
+        let payload = br#"{"ok":true}"#;
+        let mut encoder =
+            flate2::write::ZlibEncoder::new(Vec::new(), flate2::Compression::default());
+        std::io::Write::write_all(&mut encoder, payload).unwrap();
+        let compressed = encoder.finish().unwrap();
+
+        let decompressed = decompress_body("deflate", &compressed).unwrap().unwrap();
+        assert_eq!(decompressed, payload);
+    }
+
+    #[test]
+    fn decompress_body_deflate_falls_back_to_raw_stream() {
+        // 部分来源违规发 raw deflate 流，保持兼容
+        let payload = br#"{"ok":true}"#;
+        let mut encoder =
+            flate2::write::DeflateEncoder::new(Vec::new(), flate2::Compression::default());
+        std::io::Write::write_all(&mut encoder, payload).unwrap();
+        let compressed = encoder.finish().unwrap();
+
+        let decompressed = decompress_body("deflate", &compressed).unwrap().unwrap();
+        assert_eq!(decompressed, payload);
+    }
+
+    #[test]
+    fn decompress_body_zstd_roundtrip() {
+        // Codex 登录态发的就是 zstd 压缩请求体
+        let payload = br#"{"hello":"world","n":42}"#;
+        let compressed = zstd::stream::encode_all(std::io::Cursor::new(&payload[..]), 0).unwrap();
+        let decompressed = decompress_body("zstd", &compressed).unwrap().unwrap();
+        assert_eq!(decompressed, payload);
+    }
+
+    #[test]
+    fn decompress_body_stacked_gzip_then_zstd_decodes_in_reverse() {
+        // Content-Encoding: gzip, zstd 表示先 gzip 后 zstd，解压须反向（先 zstd 后 gzip）
+        let payload = br#"{"stacked":true}"#;
+        let mut gz = flate2::write::GzEncoder::new(Vec::new(), flate2::Compression::default());
+        std::io::Write::write_all(&mut gz, payload).unwrap();
+        let gzipped = gz.finish().unwrap();
+        let stacked = zstd::stream::encode_all(std::io::Cursor::new(&gzipped[..]), 0).unwrap();
+
+        let decompressed = decompress_body("gzip, zstd", &stacked).unwrap().unwrap();
+        assert_eq!(decompressed, payload);
+    }
+
+    #[test]
+    fn decompress_body_stacked_with_unsupported_returns_none() {
+        // 堆叠里只要有一个不支持，就整体保头透传
+        let result = decompress_body("snappy, zstd", b"\x00\x01\x02\x03").unwrap();
+        assert!(result.is_none());
+    }
+
+    #[test]
+    fn decompress_body_unknown_encoding_returns_none_to_keep_headers() {
+        // 未知编码必须返回 None（而非伪装成"已解码"），否则 content-encoding
+        // 头被剥掉，下游诊断会把压缩字节误报成明文
+        let result = decompress_body("snappy", b"\x00\x01\x02\x03").unwrap();
+        assert!(result.is_none());
+    }
+
+    #[test]
+    fn is_supported_content_encoding_matches_decompressable() {
+        for enc in [
+            "gzip",
+            "x-gzip",
+            "deflate",
+            "br",
+            "zstd",
+            "zst",
+            "gzip, zstd",
+        ] {
+            assert!(is_supported_content_encoding(enc), "{enc} 应受支持");
+        }
+        for enc in ["identity", "snappy", "compress", "", "gzip, snappy"] {
+            assert!(!is_supported_content_encoding(enc), "{enc} 不应受支持");
+        }
+    }
+
+    #[test]
+    fn get_content_encoding_combines_repeated_headers() {
+        // 重复的 content-encoding 头等同逗号拼接，须用 get_all 合并
+        let mut headers = HeaderMap::new();
+        headers.append("content-encoding", HeaderValue::from_static("gzip"));
+        headers.append("content-encoding", HeaderValue::from_static("zstd"));
+        assert_eq!(
+            get_content_encoding(&headers).as_deref(),
+            Some("gzip, zstd")
+        );
+    }
+
+    #[test]
+    fn get_content_encoding_ignores_identity_only() {
+        let mut headers = HeaderMap::new();
+        headers.append("content-encoding", HeaderValue::from_static("identity"));
+        assert_eq!(get_content_encoding(&headers), None);
+    }
+}
diff --git a/src-tauri/src/proxy/forwarder.rs b/src-tauri/src/proxy/forwarder.rs
index 0c840c399..d11e52901 100644
--- a/src-tauri/src/proxy/forwarder.rs
+++ b/src-tauri/src/proxy/forwarder.rs
@@ -5,6 +5,7 @@
 use super::hyper_client::ProxyResponse;
 use super::{
     body_filter::filter_private_params_with_whitelist,
+    content_encoding::{decompress_body, get_content_encoding},
     error::*,
     failover_switch::FailoverSwitchManager,
     json_canonical::{canonicalize_value, short_value_hash},
@@ -1965,7 +1966,20 @@ impl RequestForwarder {
             Ok((response, resolved_claude_api_format, outbound_model))
         } else {
             let status_code = status.as_u16();
-            let body_text = String::from_utf8(response.bytes().await?.to_vec()).ok();
+            // 错误响应同样可能被上游压缩（content-encoding）。reqwest 未启用任何
+            // 自动解压 feature，这里拿到的是原始字节；不解压的话，压缩过的错误体会
+            // 在 from_utf8 处变成非 UTF-8 而被丢弃，隐藏掉上游的限流/鉴权等详情。
+            let encoding = get_content_encoding(response.headers());
+            let raw = response.bytes().await?;
+            let decoded = match encoding {
+                Some(encoding) => match decompress_body(&encoding, &raw) {
+                    Ok(Some(decompressed)) => decompressed,
+                    // 不支持的编码 / 解压失败：退回原始字节，尽量保留可读信息
+                    _ => raw.to_vec(),
+                },
+                None => raw.to_vec(),
+            };
+            let body_text = String::from_utf8(decoded).ok();
 
             Err(ProxyError::UpstreamError {
                 status: status_code,
diff --git a/src-tauri/src/proxy/handlers.rs b/src-tauri/src/proxy/handlers.rs
index c399f1ba0..35c3062ee 100644
--- a/src-tauri/src/proxy/handlers.rs
+++ b/src-tauri/src/proxy/handlers.rs
@@ -8,6 +8,7 @@
 //! - Claude 的格式转换逻辑保留在此文件（用于 OpenRouter 旧接口回退）
 
 use super::{
+    content_encoding::{decompress_body, get_content_encoding, is_supported_content_encoding},
     error_mapper::{get_error_message, map_proxy_error_to_status},
     forwarder::ActiveConnectionGuard,
     handler_config::{
@@ -568,6 +569,49 @@ fn endpoint_with_query(uri: &axum::http::Uri, endpoint: &str) -> String {
     }
 }
 
+/// Codex 客户端（尤其 Desktop 登录态）可能对请求体启用 zstd 压缩，使得后续
+/// `serde_json::from_slice` 直接解析失败。这里在解析前解压，并剥掉已失真的实体头
+/// （content-encoding / content-length / transfer-encoding）——转发层会基于解压后的
+/// 明文 JSON 重新生成正确的头。
+fn decode_codex_request_body(
+    headers: &mut axum::http::HeaderMap,
+    body_bytes: Bytes,
+) -> Result<Bytes, ProxyError> {
+    let Some(encoding) = get_content_encoding(headers) else {
+        return Ok(body_bytes);
+    };
+
+    if !is_supported_content_encoding(&encoding) {
+        return Err(ProxyError::InvalidRequest(format!(
+            "Unsupported request content-encoding: {encoding}"
+        )));
+    }
+
+    log::debug!("[Codex] 解压请求体: content-encoding={encoding}");
+    let decompressed = match decompress_body(&encoding, &body_bytes) {
+        Ok(Some(decompressed)) => decompressed,
+        // is_supported_content_encoding 已确保编码受支持，正常不会返回 None；
+        // 防御性兜底：宁可报错，也不能把压缩字节当 JSON 透传下去。
+        Ok(None) => {
+            return Err(ProxyError::InvalidRequest(format!(
+                "Unsupported request content-encoding: {encoding}"
+            )));
+        }
+        Err(e) => {
+            log::warn!("[Codex] 请求体解压失败 ({encoding}): {e}");
+            return Err(ProxyError::InvalidRequest(format!(
+                "Failed to decompress request body ({encoding}): {e}"
+            )));
+        }
+    };
+
+    headers.remove(axum::http::header::CONTENT_ENCODING);
+    headers.remove(axum::http::header::CONTENT_LENGTH);
+    headers.remove(axum::http::header::TRANSFER_ENCODING);
+
+    Ok(Bytes::from(decompressed))
+}
+
 // ============================================================================
 // Codex API 处理器
 // ============================================================================
@@ -580,13 +624,14 @@ pub async fn handle_chat_completions(
     let (parts, req_body) = request.into_parts();
     let method = parts.method.clone();
     let uri = parts.uri;
-    let headers = parts.headers;
+    let mut headers = parts.headers;
     let extensions = parts.extensions;
     let body_bytes = req_body
         .collect()
         .await
         .map_err(|e| ProxyError::Internal(format!("Failed to read request body: {e}")))?
         .to_bytes();
+    let body_bytes = decode_codex_request_body(&mut headers, body_bytes)?;
     let body: Value = serde_json::from_slice(&body_bytes)
         .map_err(|e| ProxyError::Internal(format!("Failed to parse request body: {e}")))?;
 
@@ -645,13 +690,14 @@ pub async fn handle_responses(
     let (parts, req_body) = request.into_parts();
     let method = parts.method.clone();
     let uri = parts.uri;
-    let headers = parts.headers;
+    let mut headers = parts.headers;
     let extensions = parts.extensions;
     let body_bytes = req_body
         .collect()
         .await
         .map_err(|e| ProxyError::Internal(format!("Failed to read request body: {e}")))?
         .to_bytes();
+    let body_bytes = decode_codex_request_body(&mut headers, body_bytes)?;
     let body: Value = serde_json::from_slice(&body_bytes)
         .map_err(|e| ProxyError::Internal(format!("Failed to parse request body: {e}")))?;
 
@@ -723,13 +769,14 @@ pub async fn handle_responses_compact(
     let (parts, req_body) = request.into_parts();
     let method = parts.method.clone();
     let uri = parts.uri;
-    let headers = parts.headers;
+    let mut headers = parts.headers;
     let extensions = parts.extensions;
     let body_bytes = req_body
         .collect()
         .await
         .map_err(|e| ProxyError::Internal(format!("Failed to read request body: {e}")))?
         .to_bytes();
+    let body_bytes = decode_codex_request_body(&mut headers, body_bytes)?;
     let body: Value = serde_json::from_slice(&body_bytes)
         .map_err(|e| ProxyError::Internal(format!("Failed to parse request body: {e}")))?;
 
diff --git a/src-tauri/src/proxy/http_client.rs b/src-tauri/src/proxy/http_client.rs
index 1d3ddb2c9..97e927e46 100644
--- a/src-tauri/src/proxy/http_client.rs
+++ b/src-tauri/src/proxy/http_client.rs
@@ -223,7 +223,8 @@ fn build_client(proxy_url: Option<&str>) -> Result<Client, String> {
         // 响应解压由 response_processor 根据 content-encoding 手动处理。
         .no_gzip()
         .no_brotli()
-        .no_deflate();
+        .no_deflate()
+        .no_zstd();
 
     // 有代理地址则使用代理，否则跟随系统代理
     if let Some(url) = proxy_url {
diff --git a/src-tauri/src/proxy/mod.rs b/src-tauri/src/proxy/mod.rs
index 585703c63..cbc24404c 100644
--- a/src-tauri/src/proxy/mod.rs
+++ b/src-tauri/src/proxy/mod.rs
@@ -5,6 +5,7 @@
 pub mod body_filter;
 pub mod cache_injector;
 pub mod circuit_breaker;
+pub(crate) mod content_encoding;
 pub mod copilot_optimizer;
 pub mod error;
 pub mod error_mapper;
diff --git a/src-tauri/src/proxy/response_processor.rs b/src-tauri/src/proxy/response_processor.rs
index 59d525b1b..238e14420 100644
--- a/src-tauri/src/proxy/response_processor.rs
+++ b/src-tauri/src/proxy/response_processor.rs
@@ -3,6 +3,7 @@
 //! 统一处理流式和非流式 API 响应
 
 use super::{
+    content_encoding::{decompress_body, get_content_encoding},
     forwarder::ActiveConnectionGuard,
     handler_config::{StreamUsageEventFilter, UsageParserConfig},
     handler_context::{RequestContext, StreamingTimeoutConfig},
@@ -19,7 +20,6 @@ use bytes::Bytes;
 use futures::stream::{Stream, StreamExt};
 use serde_json::Value;
 use std::{
-    io::Read,
     sync::{
         atomic::{AtomicBool, Ordering},
         Arc,
@@ -29,60 +29,9 @@ use std::{
 use tokio::sync::Mutex;
 
 // ============================================================================
-// 响应解压
+// 响应头处理
 // ============================================================================
 
-/// 根据 content-encoding 解压响应体字节
-///
-/// reqwest 自动解压已禁用（为了透传 accept-encoding），需要手动解压。
-/// 返回 `Ok(None)` 表示编码不受支持、原样透传——此时调用方必须保留
-/// content-encoding 头，否则下游（诊断/客户端）会把压缩字节误当明文。
-fn decompress_body(content_encoding: &str, body: &[u8]) -> Result<Option<Vec<u8>>, std::io::Error> {
-    match content_encoding {
-        "gzip" | "x-gzip" => {
-            let mut decoder = flate2::read::GzDecoder::new(body);
-            let mut decompressed = Vec::new();
-            decoder.read_to_end(&mut decompressed)?;
-            Ok(Some(decompressed))
-        }
-        "deflate" => {
-            // RFC 9110: deflate 指 zlib 包裹格式；但部分上游发 raw deflate 流。
-            // 先按规范尝试 zlib，失败再回退 raw —— 否则合规上游必然解压失败，
-            // 原始压缩字节会被 fail-open 透传给 JSON 解析（#2234 形态 C 之一）。
-            let mut decompressed = Vec::new();
-            let mut zlib = flate2::read::ZlibDecoder::new(body);
-            match zlib.read_to_end(&mut decompressed) {
-                Ok(_) => Ok(Some(decompressed)),
-                Err(zlib_err) => {
-                    log::debug!("deflate 按 zlib 解压失败（{zlib_err}），回退 raw deflate");
-                    let mut decompressed = Vec::new();
-                    let mut raw = flate2::read::DeflateDecoder::new(body);
-                    raw.read_to_end(&mut decompressed)?;
-                    Ok(Some(decompressed))
-                }
-            }
-        }
-        "br" => {
-            let mut decompressed = Vec::new();
-            brotli::BrotliDecompress(&mut std::io::Cursor::new(body), &mut decompressed)?;
-            Ok(Some(decompressed))
-        }
-        _ => {
-            log::warn!("未知的 content-encoding: {content_encoding}，跳过解压");
-            Ok(None)
-        }
-    }
-}
-
-/// 从响应头提取 content-encoding（忽略 identity 和 chunked）
-fn get_content_encoding(headers: &HeaderMap) -> Option<String> {
-    headers
-        .get("content-encoding")
-        .and_then(|v| v.to_str().ok())
-        .map(|s| s.trim().to_lowercase())
-        .filter(|s| !s.is_empty() && s != "identity")
-}
-
 /// RFC 2616 / RFC 7230 中定义的不应被代理继续转发的响应头。
 const HOP_BY_HOP_RESPONSE_HEADERS: &[&str] = &[
     "connection",
@@ -878,40 +827,6 @@ mod tests {
     use std::sync::Arc;
     use tokio::sync::RwLock;
 
-    #[test]
-    fn decompress_body_deflate_handles_zlib_wrapped_per_rfc9110() {
-        // RFC 9110 规范的 deflate = zlib 包裹格式（合规上游发的就是这个）
-        let payload = br#"{"ok":true}"#;
-        let mut encoder =
-            flate2::write::ZlibEncoder::new(Vec::new(), flate2::Compression::default());
-        std::io::Write::write_all(&mut encoder, payload).unwrap();
-        let compressed = encoder.finish().unwrap();
-
-        let decompressed = decompress_body("deflate", &compressed).unwrap().unwrap();
-        assert_eq!(decompressed, payload);
-    }
-
-    #[test]
-    fn decompress_body_deflate_falls_back_to_raw_stream() {
-        // 部分上游违规发 raw deflate 流，保持兼容
-        let payload = br#"{"ok":true}"#;
-        let mut encoder =
-            flate2::write::DeflateEncoder::new(Vec::new(), flate2::Compression::default());
-        std::io::Write::write_all(&mut encoder, payload).unwrap();
-        let compressed = encoder.finish().unwrap();
-
-        let decompressed = decompress_body("deflate", &compressed).unwrap().unwrap();
-        assert_eq!(decompressed, payload);
-    }
-
-    #[test]
-    fn decompress_body_unknown_encoding_returns_none_to_keep_headers() {
-        // 未知编码必须返回 None（而非伪装成"已解码"），否则 content-encoding
-        // 头被剥掉，下游诊断会把压缩字节误报成明文
-        let result = decompress_body("zstd", b"\x28\xb5\x2f\xfd").unwrap();
-        assert!(result.is_none());
-    }
-
     #[test]
     fn test_strip_sse_field_accepts_optional_space() {
         assert_eq!(