diff --git a/frontend/resources/wasm-playground/BENCHMARKS.md b/frontend/resources/wasm-playground/BENCHMARKS.md
index 9d3ce5bce9..464de2db5f 100644
--- a/frontend/resources/wasm-playground/BENCHMARKS.md
+++ b/frontend/resources/wasm-playground/BENCHMARKS.md
@@ -16,6 +16,7 @@ it does not move caches or download missing crates.
 pnpm run benchmark:renderer run --output /tmp/baseline.json
 pnpm run benchmark:renderer run --filter texts --seed 42 --output /tmp/texts.json
 pnpm run benchmark:renderer run --headed --filter rects/default/zoom
+pnpm run benchmark:renderer run --no-build --filter rects/default/pan
 pnpm run benchmark:renderer compare /tmp/baseline.json /tmp/candidate.json
 pnpm run benchmark:renderer compare /tmp/baseline.json /tmp/candidate.json --diagnostic
 pnpm run test:renderer-benchmarks
@@ -26,7 +27,12 @@ Each run builds the ordinary optimized frontend renderer once using
 skipped, package-manager network fallback is disabled, and the emitted module is
 named `render-wasm-benchmark`. Playground assets are refreshed before launch. A
 per-run marker checks that the server points at this checkout. Concurrent builds
-or benchmark commands against the same checkout are unsupported.
+or benchmark commands against the same checkout are unsupported. `run --no-build`
+skips the build, the Cargo manifest read, and the Emscripten/esbuild preflight
+checks, then measures the existing `render-wasm-benchmark.js`/`.wasm` pair; it
+still refreshes playground assets and checks the marker. It rejects `--features`
+and never applies to `ab`, and records `metadata.build.built: false` with a
+warning that the artifact was not rebuilt from this revision.
 
 The default server URL is `http://localhost:3000`; override it with `--base-url`.
 Default JSON output goes to a unique file under the system temporary directory.
@@ -146,9 +152,11 @@ browser; the printed plan states this cost. `--output` for `ab` is a directory
 root that must not exist yet; `configs/<slug>.json`,
 `comparisons/<from>__<to>.json`, and `matrix.json` land beneath it.
 
-Run JSON is `schemaVersion` 2: `metadata.build.features` is the explicit
-sorted set (the A/B identity axis), `defaultFeatures` the resolved manifest
-defaults, plus the runner-controlled `env` map and the allowlisted
+Run JSON is `schemaVersion` 2: `metadata.build.built` is `false` when the run
+reused an existing artifact, and the build-derived fields below are absent, so
+comparing it with a built run fails compatibility. `metadata.build.features` is
+the explicit sorted set (the A/B identity axis), `defaultFeatures` the resolved
+manifest defaults, plus the runner-controlled `env` map and the allowlisted
 `ambientEnv` (`RUSTFLAGS`, `RUSTC_WRAPPER`, `CARGO_TARGET_DIR`, `CARGO_HOME`,
 `CC`, `CFLAGS`, `CXXFLAGS`, `LDFLAGS`, `PENPOT_WASM_FUNCTION_NAMES`,
 `CARGO_PROFILE_*`, `EMSDK*`). Comparison of the six build fields is always
diff --git a/frontend/scripts/renderer-benchmarks.js b/frontend/scripts/renderer-benchmarks.js
index 109820f518..a23f5aae19 100644
--- a/frontend/scripts/renderer-benchmarks.js
+++ b/frontend/scripts/renderer-benchmarks.js
@@ -59,6 +59,7 @@ export function readOptions(argv) {
       "max-features": { type: "string", default: "3" },
       confirm: { type: "boolean", default: false },
       "dry-run": { type: "boolean", default: false },
+      "no-build": { type: "boolean", default: false },
     },
   });
   const command = positionals[0] ?? "run";
@@ -89,6 +90,10 @@ export function readOptions(argv) {
   const maxFeatures = number("max-features", 1, 6);
   if (command !== "ab" && values["dry-run"])
     throw new Error("--dry-run is only valid for ab");
+  if (values["no-build"] && command !== "run")
+    throw new Error("--no-build is only valid for run");
+  if (values["no-build"] && features.length)
+    throw new Error("--no-build skips the build, so --features cannot apply");
   if (command === "ab" && !values.help) {
     if (features.some((name) => /^(stats|profile)(-|$)/.test(name)))
       throw new Error(
@@ -127,6 +132,7 @@ export function readOptions(argv) {
     maxFeatures,
     confirm: values.confirm,
     dryRun: values["dry-run"],
+    noBuild: values["no-build"],
   };
 }
 
@@ -305,14 +311,25 @@ export function recordAmbientEnv(env = process.env) {
 }
 
 async function preflight(options) {
-  await requireFile(
-    "/opt/emsdk/emsdk_env.sh",
-    "Run inside the prepared Penpot development container (Emscripten is missing).",
-  );
-  await requireFile(
-    path.join(root, "render-wasm/node_modules/esbuild/package.json"),
-    "Prepare renderer dependencies with pnpm install in render-wasm first.",
-  );
+  if (options.noBuild) {
+    for (const artifact of [
+      "render-wasm-benchmark.js",
+      "render-wasm-benchmark.wasm",
+    ])
+      await requireFile(
+        path.join(frontend, "resources/public/js", artifact),
+        "No existing renderer benchmark build; run once without --no-build first.",
+      );
+  } else {
+    await requireFile(
+      "/opt/emsdk/emsdk_env.sh",
+      "Run inside the prepared Penpot development container (Emscripten is missing).",
+    );
+    await requireFile(
+      path.join(root, "render-wasm/node_modules/esbuild/package.json"),
+      "Prepare renderer dependencies with pnpm install in render-wasm first.",
+    );
+  }
   const { chromium } = await import("playwright");
   await requireFile(
     chromium.executablePath(),
@@ -389,6 +406,7 @@ export async function run(options) {
       },
       build: {
         mode: "release",
+        built: !options.noBuild,
         featurePolicy: "manifest-defaults",
         features: [],
         target: "frontend",
@@ -474,34 +492,41 @@ export async function run(options) {
       ),
     );
     result.metadata.environment.playwright = packageJson.version;
-    const requested = [...(options.features ?? [])].sort();
-    const resolved = validateFeatureSelection(
-      readRendererManifest(),
-      requested,
-      options.command,
-    );
-    result.metadata.build.features = requested;
-    result.metadata.build.defaultFeatures = resolved;
-    if (requested.length) result.metadata.build.featurePolicy = "explicit";
-    result.metadata.build.env = { ...BUILD_ENV };
-    result.metadata.build.ambientEnv = recordAmbientEnv();
-    // Disallow package-manager network fallback in a prepared run.
-    const buildStart = performance.now();
-    await execute(
-      "./build",
-      [
-        "frontend",
-        "--offline",
-        "--locked",
-        ...(requested.length ? ["--features", requested.join(",")] : []),
-      ],
-      {
-        cwd: path.join(root, "render-wasm"),
-        signal: abort.signal,
-        env: { ...process.env, ...BUILD_ENV, VERSION: runId },
-      },
-    );
-    result.metadata.build.durationMs = performance.now() - buildStart;
+    if (options.noBuild) {
+      const warning =
+        "Build skipped (--no-build): the measured artifact was not rebuilt from this revision.";
+      result.warnings.push(warning);
+      console.warn(warning);
+    } else {
+      const requested = [...(options.features ?? [])].sort();
+      const resolved = validateFeatureSelection(
+        readRendererManifest(),
+        requested,
+        options.command,
+      );
+      result.metadata.build.features = requested;
+      result.metadata.build.defaultFeatures = resolved;
+      if (requested.length) result.metadata.build.featurePolicy = "explicit";
+      result.metadata.build.env = { ...BUILD_ENV };
+      result.metadata.build.ambientEnv = recordAmbientEnv();
+      // Disallow package-manager network fallback in a prepared run.
+      const buildStart = performance.now();
+      await execute(
+        "./build",
+        [
+          "frontend",
+          "--offline",
+          "--locked",
+          ...(requested.length ? ["--features", requested.join(",")] : []),
+        ],
+        {
+          cwd: path.join(root, "render-wasm"),
+          signal: abort.signal,
+          env: { ...process.env, ...BUILD_ENV, VERSION: runId },
+        },
+      );
+      result.metadata.build.durationMs = performance.now() - buildStart;
+    }
     const { copyWasmPlayground } = await import("./_helpers.js");
     const previousCwd = process.cwd();
     try {
@@ -902,7 +927,7 @@ export async function main(argv = process.argv.slice(2)) {
   const options = readOptions(argv);
   if (options.help) {
     console.log(
-      "Renderer benchmarks\n  node scripts/renderer-benchmarks.js run [--filter ID] [--seed N] [--warmups N] [--repetitions N] [--timeout-ms N] [--headed] [--base-url URL] [--features a,b] [--output FILE]\n  node scripts/renderer-benchmarks.js ab --features a,b [--max-features N] [--confirm] [--dry-run] [--filter ID] [--seed N] [--warmups N] [--repetitions N] [--timeout-ms N] [--headed] [--base-url URL] [--output DIR]\n  node scripts/renderer-benchmarks.js compare BASELINE CANDIDATE [--diagnostic] [--output FILE]\nRun requires a prepared development container and an existing static server on port 3000. Defaults: 3 warmups, 10 measured attempts, 30s timeout; provisional pending calibration.\n" +
+      "Renderer benchmarks\n  node scripts/renderer-benchmarks.js run [--filter ID] [--seed N] [--warmups N] [--repetitions N] [--timeout-ms N] [--headed] [--base-url URL] [--features a,b] [--output FILE] [--no-build]\n  node scripts/renderer-benchmarks.js ab --features a,b [--max-features N] [--confirm] [--dry-run] [--filter ID] [--seed N] [--warmups N] [--repetitions N] [--timeout-ms N] [--headed] [--base-url URL] [--output DIR]\n  node scripts/renderer-benchmarks.js compare BASELINE CANDIDATE [--diagnostic] [--output FILE]\nRun requires a prepared development container and an existing static server on port 3000. --no-build reuses the existing renderer artifact instead of building and cannot be combined with --features or ab. Defaults: 3 warmups, 10 measured attempts, 30s timeout; provisional pending calibration.\n" +
         LIMITS,
     );
     return 0;
diff --git a/frontend/scripts/renderer-benchmarks/tests/runner.test.js b/frontend/scripts/renderer-benchmarks/tests/runner.test.js
index cdc8293d88..86320f30bc 100644
--- a/frontend/scripts/renderer-benchmarks/tests/runner.test.js
+++ b/frontend/scripts/renderer-benchmarks/tests/runner.test.js
@@ -137,6 +137,18 @@ test("CLI parses feature selection for run and ab", () => {
   }
 });
 
+test("--no-build skips the renderer build only for run", () => {
+  assert.equal(readOptions(["run"]).noBuild, false);
+  assert.equal(readOptions(["run", "--no-build"]).noBuild, true);
+  for (const args of [
+    ["ab", "--features", "branch-b", "--no-build"],
+    ["compare", "a.json", "b.json", "--no-build"],
+    ["run", "--no-build", "--features", "branch-b"],
+  ]) {
+    assert.throws(() => readOptions(args), /--no-build/);
+  }
+});
+
 test("feature selection validates against the renderer manifest", () => {
   const definitions = { default: [], "branch-a": [], "branch-b": [] };
   assert.deepEqual(validateFeatureSelection(definitions, ["branch-b"], "ab"), [