diff --git a/CHANGELOG.md b/CHANGELOG.md
index c56f7a43b2..e91485edab 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -21,6 +21,7 @@ _Living section — bullets land here as PRs merge into `release/v3.8.47` (paral
 - **fix(resilience):** a bare, unrecognized `403` from a no-credential (`authType:"none"`) provider like mimocode or theoldllm no longer permanently bans the connection ([#6315](https://github.com/diegosouzapw/OmniRoute/issues/6315), [#6345](https://github.com/diegosouzapw/OmniRoute/issues/6345)) — `classifyProviderError()`'s 403 branch only exempted `apikey` providers from the terminal `FORBIDDEN` classification, so these free/stateless proxies (no real account/credential to revoke) fell through to `FORBIDDEN` on the first unmatched 403 and got `isActive:false, testStatus:"banned"` with no cooldown or retry. The exemption now also covers `authType:"none"` providers, returning `null` (recoverable) so the existing connection-cooldown/retry layer handles it. Regression guard: `tests/unit/errorClassifier-noauth-403-6315.test.ts`.
 - **fix(providers):** the **Auggie (Augment CLI)** executor no longer fails on Windows with `spawn EINVAL` ([#6304](https://github.com/diegosouzapw/OmniRoute/issues/6304)) — the global-npm install exposes `auggie` as a `.cmd` shim, which Node's `child_process.spawn` cannot launch on win32 without `shell: true`. Both spawn sites (streaming + the `auggie --version` test) now go through a shared `buildAuggieSpawnOptions()` that sets `shell: process.platform === "win32"`; the argv (built by `buildAuggieArgs()` with a registry-validated `model` and a trailing `--` end-of-options marker) is unchanged, so the argument-injection surface stays closed on non-Windows. Regression guard: `tests/unit/auggie-win32-spawn-6304.test.ts`.
 - **fix(api):** the dashboard **"Test model"** action is now a clean connection test ([#6240](https://github.com/diegosouzapw/OmniRoute/issues/6240)) — `modelTestRunner` sent its probe request without an explicit compression override, so whenever the operator's global `compression.enabled` flag was on the test call inherited compression (and any Output-Styles system prompt), polluting the result. The internal test requests now send `X-OmniRoute-Compression: off`, and `chatCore` honors an explicit `off` header even when `compression.enabled` is globally true. Regression guards: `tests/unit/model-test-runner-compression-off-6240.test.ts`, `tests/integration/test-model-compression-off-6240.test.ts`.
+- **fix(api):** `HEAD` requests no longer hang until client timeout on any route — valid, unknown, authed, or unauthed ([#6400](https://github.com/diegosouzapw/OmniRoute/issues/6400)), broader follow-up to the route-specific #6517 (`/v1/models`). Root cause: Next.js 16's App Router *route-handler* pipeline (`next/dist/server/send-response.js`) correctly skips piping a `Response` body for `HEAD`, but its *page*-rendering pipeline (`next/dist/server/pipe-readable.js` → `pipeToNodeResponse`, used for every app-router page/layout render — including the `not-found` boundary any unmatched path falls through to) has no such check and always streams the full rendered body regardless of method; combined with Node's default keep-alive framing this left some clients unsure whether the (implicitly bodyless) `HEAD` response had actually finished. A new `scripts/dev/head-response-guard.cjs`, wired into both the dev/start custom server (`scripts/dev/run-next.mjs`) and the packaged standalone server (`scripts/dev/standalone-server-ws.mjs`) at the same tier as the existing `http-method-guard.cjs`/`peer-stamp.mjs` wrappers, discards any body bytes written for a `HEAD` request and forces `Connection: close` once `.end()` is called — independent of route existence or auth state, satisfying RFC 9110 §9.3.2. Regression guard: `tests/unit/head-request-closes-6400.test.ts`.
 
 ### 📝 Maintenance
 
diff --git a/scripts/dev/head-response-guard.cjs b/scripts/dev/head-response-guard.cjs
new file mode 100644
index 0000000000..7ecb460439
--- /dev/null
+++ b/scripts/dev/head-response-guard.cjs
@@ -0,0 +1,104 @@
+"use strict";
+
+/**
+ * HEAD response guard (#6400).
+ *
+ * RFC 9110 §9.3.2 requires a HEAD response to carry the same headers/status a
+ * GET would, with ZERO body, and the connection should not leave the client
+ * guessing about when the (bodyless) response is actually finished.
+ *
+ * Next.js 16 handles this correctly for App Router *route handlers*
+ * (`route.ts` exporting `GET`) — `next/dist/server/send-response.js` explicitly
+ * skips piping the `Response.body` when `req.method === 'HEAD'`. But Next's
+ * *page*-rendering pipeline (`next/dist/server/pipe-readable.js` ->
+ * `pipeToNodeResponse`, used for every app-router page/layout render — the
+ * root page, the `not-found` boundary that unmatched paths fall through to,
+ * dashboard pages, etc.) has NO such check: it always pipes the fully
+ * rendered body to the HTTP response regardless of method. Combined with
+ * Node's default keep-alive framing, a HEAD request to any page-rendered path
+ * ends up with the socket only settling once that render finishes — on a
+ * client that doesn't special-case a HEAD response's implicit zero-length
+ * body (observed on Windows/curl in #6400), this reads as "headers arrive,
+ * then it hangs" instead of the RFC-mandated "closes immediately".
+ *
+ * Fix: for every inbound HEAD request, before Next ever sees it, wrap the
+ * Node `ServerResponse` so:
+ *   - Any body bytes written by Next (route handler OR page render) are
+ *     discarded — status code and headers Next computed (auth 401s, 404s,
+ *     200s, etc.) are preserved untouched.
+ *   - The connection is force-closed right after headers flush
+ *     (`Connection: close`), removing any keep-alive ambiguity a client could
+ *     have about whether more bytes are coming.
+ *
+ * This applies globally (valid routes, unmatched/404 paths, authed and
+ * unauthed) because it operates at the Node HTTP transport layer shared by
+ * every request — the same tier as the existing `http-method-guard.cjs` /
+ * `peer-stamp.mjs` wrappers — never inside Next's per-route code.
+ * See: https://github.com/diegosouzapw/OmniRoute/issues/6400
+ */
+
+function isHeadRequest(req) {
+  return typeof req?.method === "string" && req.method.toUpperCase() === "HEAD";
+}
+
+/**
+ * Mutates `res` in place so any body write is discarded and the response
+ * ends (closing the connection) as soon as `.end()` is called, regardless of
+ * what body argument was passed to it.
+ *
+ * @param {import("node:http").ServerResponse} res
+ */
+function suppressBodyAndForceClose(res) {
+  try {
+    // Never leave the client guessing whether the (bodyless) response has
+    // more bytes coming — closing the socket is the unambiguous signal.
+    res.setHeader("Connection", "close");
+  } catch {
+    // Headers may already be flushed in rare re-entrant cases — the write/end
+    // overrides below still guarantee an empty, prompt HEAD response.
+  }
+
+  const originalEnd = res.end.bind(res);
+  let ended = false;
+
+  res.write = function headSuppressedWrite(_chunk, encodingOrCb, cb) {
+    // Discard the body but keep the writable-stream contract: report the
+    // write as flushed (no backpressure) so callers like Next's
+    // `pipeToNodeResponse` never block waiting on a `drain` that would
+    // otherwise never fire, and invoke whichever callback form was passed.
+    if (typeof encodingOrCb === "function") encodingOrCb();
+    else if (typeof cb === "function") cb();
+    return true;
+  };
+
+  res.end = function headSuppressedEnd(chunk, encoding, cb) {
+    if (ended) return res;
+    ended = true;
+    if (typeof chunk === "function") return originalEnd(chunk);
+    if (typeof encoding === "function") return originalEnd(encoding);
+    if (typeof cb === "function") return originalEnd(cb);
+    return originalEnd();
+  };
+}
+
+/**
+ * Wrap a Node request listener so every inbound HEAD request gets the
+ * body-suppression + forced-close treatment before the wrapped listener
+ * (eventually Next.js) runs.
+ *
+ * @param {(req: import("node:http").IncomingMessage, res: import("node:http").ServerResponse) => unknown} listener
+ */
+function wrapRequestListenerWithHeadResponseGuard(listener) {
+  return function headResponseGuardRequestHandler(req, res) {
+    if (isHeadRequest(req)) {
+      suppressBodyAndForceClose(res);
+    }
+    return listener.call(this, req, res);
+  };
+}
+
+module.exports = {
+  isHeadRequest,
+  suppressBodyAndForceClose,
+  wrapRequestListenerWithHeadResponseGuard,
+};
diff --git a/scripts/dev/run-next.mjs b/scripts/dev/run-next.mjs
index 4a402650a8..0ff02837f0 100644
--- a/scripts/dev/run-next.mjs
+++ b/scripts/dev/run-next.mjs
@@ -10,6 +10,7 @@ import { createOmnirouteWsBridge } from "./v1-ws-bridge.mjs";
 import { createResponsesWsProxy } from "./responses-ws-proxy.mjs";
 import { ensurePeerStampToken, stampPeerIp } from "./peer-stamp.mjs";
 import methodGuard from "./http-method-guard.cjs";
+import headResponseGuard from "./head-response-guard.cjs";
 import { ensureNativeSqlite } from "./ensure-native-sqlite.mjs";
 import {
   isTurbopackCacheCorruption,
@@ -18,6 +19,7 @@ import {
 import { randomUUID } from "node:crypto";
 
 const { maybeHandleDisallowedMethod } = methodGuard;
+const { wrapRequestListenerWithHeadResponseGuard } = headResponseGuard;
 
 // Pre-read DATA_DIR from local .env before bootstrap resolves paths
 if (!process.env.DATA_DIR) {
@@ -143,13 +145,15 @@ async function start() {
     baseUrl: `http://127.0.0.1:${dashboardPort}`,
   });
 
-  const server = http.createServer((req, res) => {
-    if (maybeHandleDisallowedMethod(req, res)) return;
-    // Stamp the real TCP peer IP before Next sees the request, so the authz
-    // middleware can decide LOCAL_ONLY locality without trusting the Host header.
-    stampPeerIp(req);
-    return requestHandler(req, res);
-  });
+  const server = http.createServer(
+    wrapRequestListenerWithHeadResponseGuard((req, res) => {
+      if (maybeHandleDisallowedMethod(req, res)) return;
+      // Stamp the real TCP peer IP before Next sees the request, so the authz
+      // middleware can decide LOCAL_ONLY locality without trusting the Host header.
+      stampPeerIp(req);
+      return requestHandler(req, res);
+    })
+  );
   server.on("upgrade", async (req, socket, head) => {
     try {
       const responsesWsHandled = await responsesWsProxy.handleUpgrade(req, socket, head);
diff --git a/scripts/dev/standalone-server-ws.mjs b/scripts/dev/standalone-server-ws.mjs
index 890caee40f..eb59d9d50a 100644
--- a/scripts/dev/standalone-server-ws.mjs
+++ b/scripts/dev/standalone-server-ws.mjs
@@ -5,11 +5,13 @@ import { createResponsesWsProxy } from "./responses-ws-proxy.mjs";
 import { ensurePeerStampToken, wrapRequestListenerWithPeerStamp } from "./peer-stamp.mjs";
 import { maybeHandleWebdav } from "./webdav-handler.mjs";
 import methodGuard from "./http-method-guard.cjs";
+import headResponseGuard from "./head-response-guard.cjs";
 import { resolveTlsOptions, createServerListener } from "./tls-options.mjs";
 
 const originalCreateServer = http.createServer.bind(http);
 const proxiesByPort = new Map();
 const { wrapRequestListenerWithMethodGuard } = methodGuard;
+const { wrapRequestListenerWithHeadResponseGuard } = headResponseGuard;
 
 // Opt-in native HTTPS (#5242). Resolved once at boot: when both OMNIROUTE_TLS_CERT
 // and OMNIROUTE_TLS_KEY point at readable files we terminate TLS on the same
@@ -114,8 +116,12 @@ http.createServer = function createServerWithResponsesWs(...args) {
   const lastFnIdx = args.map((a) => typeof a === "function").lastIndexOf(true);
   if (lastFnIdx >= 0) {
     // Method guard runs before Next because Next 16 rejects TRACE while constructing requests.
-    args[lastFnIdx] = wrapRequestListenerWithMethodGuard(
-      wrapRequestListenerWithWebdav(wrapRequestListenerWithPeerStamp(args[lastFnIdx]))
+    // Head-response guard wraps outermost so it sees (and can force-close) every
+    // HEAD request regardless of which inner layer ends up handling it (#6400).
+    args[lastFnIdx] = wrapRequestListenerWithHeadResponseGuard(
+      wrapRequestListenerWithMethodGuard(
+        wrapRequestListenerWithWebdav(wrapRequestListenerWithPeerStamp(args[lastFnIdx]))
+      )
     );
   }
 
@@ -134,8 +140,10 @@ http.createServer = function createServerWithResponsesWs(...args) {
     if (eventName === "request" && typeof listener === "function") {
       return originalOn(
         eventName,
-        wrapRequestListenerWithMethodGuard(
-          wrapRequestListenerWithWebdav(wrapRequestListenerWithPeerStamp(listener))
+        wrapRequestListenerWithHeadResponseGuard(
+          wrapRequestListenerWithMethodGuard(
+            wrapRequestListenerWithWebdav(wrapRequestListenerWithPeerStamp(listener))
+          )
         )
       );
     }
@@ -149,8 +157,10 @@ http.createServer = function createServerWithResponsesWs(...args) {
     if (eventName === "request" && typeof listener === "function") {
       return originalAddListener(
         eventName,
-        wrapRequestListenerWithMethodGuard(
-          wrapRequestListenerWithWebdav(wrapRequestListenerWithPeerStamp(listener))
+        wrapRequestListenerWithHeadResponseGuard(
+          wrapRequestListenerWithMethodGuard(
+            wrapRequestListenerWithWebdav(wrapRequestListenerWithPeerStamp(listener))
+          )
         )
       );
     }
diff --git a/tests/unit/head-request-closes-6400.test.ts b/tests/unit/head-request-closes-6400.test.ts
new file mode 100644
index 0000000000..f10180911c
--- /dev/null
+++ b/tests/unit/head-request-closes-6400.test.ts
@@ -0,0 +1,237 @@
+import { describe, it, after } from "node:test";
+import assert from "node:assert/strict";
+import http from "node:http";
+import { EventEmitter } from "node:events";
+import { createRequire } from "node:module";
+import type { AddressInfo } from "node:net";
+
+const require = createRequire(import.meta.url);
+const headResponseGuard = require("../../scripts/dev/head-response-guard.cjs") as {
+  wrapRequestListenerWithHeadResponseGuard: (
+    listener: (req: http.IncomingMessage, res: http.ServerResponse) => unknown
+  ) => (req: http.IncomingMessage, res: http.ServerResponse) => unknown;
+  suppressBodyAndForceClose: (res: http.ServerResponse) => void;
+};
+
+const { wrapRequestListenerWithHeadResponseGuard, suppressBodyAndForceClose } = headResponseGuard;
+
+/**
+ * Regression test for issue #6400 — "HEAD requests hang ~6s — response never
+ * closes after headers", reported across EVERY route (valid, unknown, authed,
+ * unauthed).
+ *
+ * Root cause: Next.js 16's App Router route-handler pipeline
+ * (`next/dist/server/send-response.js`) correctly skips piping a `Response`
+ * body for HEAD requests, but its *page*-rendering pipeline
+ * (`next/dist/server/pipe-readable.js` -> `pipeToNodeResponse`, used for every
+ * app-router page/layout render — including the `not-found` boundary that
+ * unmatched paths fall through to) has NO such check: it always streams the
+ * full rendered body to the HTTP response regardless of method. Combined with
+ * Node's default keep-alive framing this leaves clients (observed on
+ * Windows/curl) unsure whether the — implicitly bodyless — HEAD response has
+ * actually finished.
+ *
+ * Fix: `scripts/dev/head-response-guard.cjs` wraps the Node request listener
+ * (wired into both the dev/start custom server `scripts/dev/run-next.mjs` and
+ * the packaged standalone server `scripts/dev/standalone-server-ws.mjs`) so
+ * that, for every inbound HEAD request, any body bytes the inner handler
+ * tries to write are discarded (never blocking on backpressure) and the
+ * connection is force-closed (`Connection: close`) as soon as `.end()` is
+ * called — independent of route existence or auth state.
+ */
+describe("issue #6400 — HEAD response guard (unit)", () => {
+  function makeMockResponse() {
+    const emitter = new EventEmitter() as EventEmitter & {
+      headers: Record<string, string>;
+      ended: boolean;
+      writeCalls: unknown[][];
+      endCalls: unknown[][];
+      write: (...args: unknown[]) => boolean;
+      end: (...args: unknown[]) => unknown;
+      setHeader: (name: string, value: string) => void;
+    };
+    emitter.headers = {};
+    emitter.ended = false;
+    emitter.writeCalls = [];
+    emitter.endCalls = [];
+    emitter.setHeader = (name: string, value: string) => {
+      emitter.headers[name.toLowerCase()] = value;
+    };
+    emitter.write = (...args: unknown[]) => {
+      emitter.writeCalls.push(args);
+      return true;
+    };
+    emitter.end = (...args: unknown[]) => {
+      emitter.endCalls.push(args);
+      emitter.ended = true;
+      return emitter;
+    };
+    return emitter;
+  }
+
+  it("sets Connection: close on the response", () => {
+    const res = makeMockResponse();
+    suppressBodyAndForceClose(res as unknown as http.ServerResponse);
+    assert.equal(res.headers.connection, "close");
+  });
+
+  it("discards any body written via res.write() but still reports success (no backpressure stall)", () => {
+    const res = makeMockResponse();
+    const originalWrite = res.write;
+    suppressBodyAndForceClose(res as unknown as http.ServerResponse);
+
+    const ok = res.write("this body must never reach the socket");
+    assert.equal(ok, true, "write must report success so callers never block on a drain event");
+    assert.equal(
+      res.writeCalls.length,
+      0,
+      "the original write() must never be called — the body must be fully discarded"
+    );
+    assert.notEqual(res.write, originalWrite);
+  });
+
+  it("res.end() forwards to the original end with NO body argument, and is idempotent", () => {
+    const res = makeMockResponse();
+    suppressBodyAndForceClose(res as unknown as http.ServerResponse);
+
+    res.end("<html>this must be dropped</html>");
+    res.end("second call must be a no-op");
+
+    assert.equal(res.endCalls.length, 1, "end() must only forward to the original once");
+    assert.deepEqual(
+      res.endCalls[0],
+      [],
+      "the discarded body must never be forwarded to the real end()"
+    );
+    assert.equal(res.ended, true);
+  });
+
+  it("wrapRequestListenerWithHeadResponseGuard only guards HEAD requests, GET/POST pass through untouched", () => {
+    let receivedReq: { method?: string } | null = null;
+    let receivedRes: unknown = null;
+    const listener = (req: { method?: string }, res: unknown) => {
+      receivedReq = req;
+      receivedRes = res;
+    };
+    const guarded = wrapRequestListenerWithHeadResponseGuard(
+      listener as unknown as (req: http.IncomingMessage, res: http.ServerResponse) => unknown
+    );
+
+    const getRes = makeMockResponse();
+    guarded({ method: "GET" } as unknown as http.IncomingMessage, getRes as unknown as http.ServerResponse);
+    assert.equal(getRes.headers.connection, undefined, "GET must not be forced to close");
+
+    const headRes = makeMockResponse();
+    guarded(
+      { method: "HEAD" } as unknown as http.IncomingMessage,
+      headRes as unknown as http.ServerResponse
+    );
+    assert.equal(headRes.headers.connection, "close", "HEAD must be force-closed");
+
+    // Sanity: the inner listener was actually invoked in both cases (guard
+    // must not swallow the request — routing/auth/status-code logic still
+    // runs, only the body write path is intercepted).
+    assert.ok(receivedReq);
+    assert.ok(receivedRes);
+  });
+});
+
+/**
+ * End-to-end confirmation over a real TCP socket: a HEAD request through the
+ * guarded listener gets an empty body and a closed connection even when the
+ * underlying handler writes a large body synchronously (the shape that, on
+ * an un-guarded pipe, streams the full page/RSC payload for HEAD exactly as
+ * Next's `pipeToNodeResponse` does today).
+ */
+describe("issue #6400 — HEAD response guard (integration, real socket)", () => {
+  const servers: http.Server[] = [];
+
+  after(() => {
+    for (const server of servers) server.close();
+  });
+
+  function startGuardedServer(
+    handler: (req: http.IncomingMessage, res: http.ServerResponse) => void
+  ): Promise<{ port: number }> {
+    const server = http.createServer(wrapRequestListenerWithHeadResponseGuard(handler));
+    servers.push(server);
+    return new Promise((resolve) => {
+      server.listen(0, "127.0.0.1", () => resolve({ port: (server.address() as AddressInfo).port }));
+    });
+  }
+
+  function rawRequest(
+    port: number,
+    method: string
+  ): Promise<{ statusCode: number; headers: Record<string, string>; body: string }> {
+    return new Promise((resolvePromise, reject) => {
+      const req = http.request(
+        { host: "127.0.0.1", port, method, path: "/", headers: { Connection: "keep-alive" } },
+        (res) => {
+          const chunks: Buffer[] = [];
+          res.on("data", (chunk) => chunks.push(chunk));
+          res.on("end", () => {
+            resolvePromise({
+              statusCode: res.statusCode ?? 0,
+              headers: Object.fromEntries(
+                Object.entries(res.headers).map(([k, v]) => [k, String(v)])
+              ),
+              body: Buffer.concat(chunks).toString("utf8"),
+            });
+          });
+        }
+      );
+      req.on("error", reject);
+      req.end();
+    });
+  }
+
+  it("HEAD to a handler that writes a large body synchronously still returns an empty body + Connection: close", async () => {
+    const largeBody = "x".repeat(1_000_000);
+    const { port } = await startGuardedServer((req, res) => {
+      res.writeHead(200, { "Content-Type": "text/html" });
+      res.write(largeBody);
+      res.end();
+    });
+
+    const result = await rawRequest(port, "HEAD");
+
+    assert.equal(result.statusCode, 200);
+    assert.equal(result.body, "", "HEAD body must be empty per RFC 9110 §9.3.2");
+    assert.equal(
+      result.headers.connection,
+      "close",
+      "HEAD response must force Connection: close so the client never has to guess"
+    );
+  });
+
+  it("GET through the SAME guarded listener is unaffected (still streams the full body)", async () => {
+    const { port } = await startGuardedServer((req, res) => {
+      res.writeHead(200, { "Content-Type": "text/plain" });
+      res.end("hello world");
+    });
+
+    const result = await rawRequest(port, "GET");
+
+    assert.equal(result.statusCode, 200);
+    assert.equal(result.body, "hello world");
+    assert.notEqual(
+      result.headers.connection,
+      "close",
+      "GET requests must not be forced to close — only HEAD is guarded"
+    );
+  });
+
+  it("HEAD to a 404/unknown-route-shaped handler also closes immediately with an empty body", async () => {
+    const { port } = await startGuardedServer((req, res) => {
+      res.writeHead(404, { "Content-Type": "text/html" });
+      res.end("<html>not found</html>");
+    });
+
+    const result = await rawRequest(port, "HEAD");
+
+    assert.equal(result.statusCode, 404);
+    assert.equal(result.body, "");
+    assert.equal(result.headers.connection, "close");
+  });
+});