diff --git a/README.md b/README.md
index 2cecdcbfbc4..27e0b33dcf2 100644
--- a/README.md
+++ b/README.md
@@ -38,7 +38,7 @@
 > | host | build | since | record |
 > |---|---|---|---|
 > | Apple Silicon, mlx-metal (`:11435`) | `0.40.2-dynres.1-0-g3960244`, both halves built at the point tag | 2026-10-13 22:01 | [#486](https://github.com/MaxusAI/ollama/pull/486) |
-> | CUDA host, ai-server (`:11497`) | `0.40.2-dynres.1-4-g377f9d1`: the point tag plus [#487](https://github.com/MaxusAI/ollama/pull/487), a Go-only build of main `377f9d14` (image `main-377f9d14`) | 2026-10-13 23:39 | [#484](https://github.com/MaxusAI/ollama/pull/484), [#487](https://github.com/MaxusAI/ollama/pull/487) |
+> | CUDA host, ai-server (`:11497`) | `0.40.2-dynres.1-17-g75c6d35`: the v0.40.3 fold ([#488](https://github.com/MaxusAI/ollama/pull/488)) on the point tag plus [#487](https://github.com/MaxusAI/ollama/pull/487), a Go-only build of main `75c6d350` (image `main-75c6d350`) | 2026-10-14 07:53 | [#487](https://github.com/MaxusAI/ollama/pull/487), [#488](https://github.com/MaxusAI/ollama/pull/488), [#492](https://github.com/MaxusAI/ollama/pull/492) |
 > | AMD gfx1151 | `v0.35.0-dynres`; the `v0.40.2-dynres.1` release build passes `rocm7-0-40-2`, and its deploy is scheduled after the [#477](https://github.com/MaxusAI/ollama/pull/477) campaign | 2026-10-01 | [#485](https://github.com/MaxusAI/ollama/pull/485) |
 > | H100 (CUDA) | `0.35.0-dynres-15-ga657392`, a build of `main` past the v0.35.0 tag | 2026-10-03 | [v0.35.0 record](docs/maxusai/tasks/upstream-sync-0.35.0.md) |
 >
@@ -77,16 +77,18 @@
        release_matrix.py --version 0.40.2-dynres.1 \
            runs/preflight-mlx-metal-0402-dynres1-prod-g3960244.json \
            runs/preflight-rocm7-0402-dynres1-release-g3960244.json \
-           runs/preflight-cuda-0402-dynres1-487-prod-g377f9d1.json \
-           runs/preflight-mlx-cuda-0402-dynres1-487-prod-g377f9d1.json
+           runs/preflight-cuda-0403-prod-g75c6d35.json \
+           runs/preflight-mlx-cuda-0403-prod-g75c6d35.json
      The runs are listed by name because runs/ also holds #487's A/B runs and canaries, which are not release runs.
-     The CUDA host serves 0.40.2-dynres.1-4-g377f9d1 (main 377f9d14: the point tag plus #487, Go only); both of its
-     rows are production's own runs on that build, with both GPUs visible. -->
+     The CUDA host serves 0.40.2-dynres.1-17-g75c6d35 (main 75c6d350: the v0.40.3 fold on the point tag plus #487, Go
+     only); both of its rows are production's own runs on that build. The cuda row is the idle re-run
+     (preflight-cuda-0403-prod-g75c6d35.json); the first post-deploy run, -contended.json, failed gemma4's ladder and
+     pinned budget by a constant +573 while a client's gemma4 q8 shared the GPU and capped image_max_tokens to n_ubatch. -->
 
 | surface | Build identity | MLX build pin | Toolchain pin | M5 tensor path | Image size ladder | Aspect ladder | Pinned image budget | thinking on/off | Output quality | fp16 overflow canary | Runner isolation | measured on |
 |---|---|---|---|---|---|---|---|---|---|---|---|---|
-| **cuda** | green | green | skipped | n/a | green | skipped | green | green | not run | skipped | green | `0.40.2-dynres.1-4-g377f9d1` |
-| **mlx-cuda** | skipped | green | skipped | n/a | green | skipped | skipped | skipped | not run | skipped | green | `0.40.2-dynres.1-4-g377f9d1` |
+| **cuda** | green | green | skipped | n/a | green | skipped | green | green | not run | skipped | green | `0.40.2-dynres.1-17-g75c6d35` |
+| **mlx-cuda** | skipped | green | skipped | n/a | green | skipped | skipped | skipped | not run | skipped | green | `0.40.2-dynres.1-17-g75c6d35` |
 | **mlx-metal** | skipped | green | skipped | green | green | green | skipped | green | skipped | skipped | green | `0.40.2-dynres.1-0-g3960244` |
 | **apple-silicon-mlx** | not run | not run | not run | not run | not run | not run | not run | not run | not run | not run | not run | — |
 | **rocm7** | green | skipped | green | n/a | green | green | green | green | green | green | green | `0.40.2-dynres.1-0-g3960244` |
diff --git a/docs/maxusai/vision-suite/preflight/runs/preflight-cuda-0403-canary-gpusall-g75c6d35.json b/docs/maxusai/vision-suite/preflight/runs/preflight-cuda-0403-canary-gpusall-g75c6d35.json
new file mode 100644
index 00000000000..7a6347a9c21
--- /dev/null
+++ b/docs/maxusai/vision-suite/preflight/runs/preflight-cuda-0403-canary-gpusall-g75c6d35.json
@@ -0,0 +1,472 @@
+{
+  "meta": {
+    "host": "http://127.0.0.1:11532",
+    "platform": "cuda",
+    "image_tag": "maxusai/ollama:main-75c6d350",
+    "started_utc": "20261013T214231",
+    "expectations_file": "expectations.toml",
+    "schema_version": 1,
+    "version": "0.40.2-dynres.1-17-g75c6d35",
+    "profile": "cuda-0-40-2",
+    "patchset": [
+      "001",
+      "002",
+      "004",
+      "005",
+      "801",
+      "802",
+      "805",
+      "903",
+      "908"
+    ],
+    "container": "ollama-ab-0402-canary0403-gpusall",
+    "llama_cpp_build": "631109b34",
+    "mlx_build": "0.32.3-53-ga59cc23"
+  },
+  "results": [
+    {
+      "check": "version",
+      "arch": null,
+      "status": "PASS",
+      "summary": "0.40.2-dynres.1-17-g75c6d35 matches profile cuda-0-40-2",
+      "expected": "^0\\.40\\.2-dynres(\\.\\d+)?-\\d+-g[0-9a-f]{7,40}$",
+      "actual": "0.40.2-dynres.1-17-g75c6d35"
+    },
+    {
+      "check": "payload_pin",
+      "arch": null,
+      "status": "PASS",
+      "summary": "llama.cpp build 631109b34 matches the measured payload (docker exec)",
+      "expected": "631109b34",
+      "actual": "631109b34"
+    },
+    {
+      "check": "toolchain_pin",
+      "arch": null,
+      "status": "SKIP",
+      "summary": "profile records no toolchain_build to assert against",
+      "diagnosis": "Add toolchain_build (e.g. \"rocm-7.2.4\" or \"cuda-13.0\") so a ROCm/CUDA bump under an unchanged payload fails loudly instead of inheriting these expectations. Platforms with no such toolchain (metal, cpu) leave it unset and keep this skip; the MLX library is pinned separately by mlx_build."
+    },
+    {
+      "check": "image_tag",
+      "arch": null,
+      "status": "PASS",
+      "summary": "ollama-ab-0402-canary0403-gpusall runs maxusai/ollama:main-75c6d350",
+      "expected": "maxusai/ollama:main-75c6d350",
+      "actual": "maxusai/ollama:main-75c6d350"
+    },
+    {
+      "check": "go_patch_marker",
+      "arch": null,
+      "status": "PASS",
+      "summary": "--image-max-tokens present in binary (1)",
+      "expected": 1,
+      "actual": 1
+    },
+    {
+      "check": "metal_tensor_host",
+      "arch": null,
+      "status": "SKIP",
+      "summary": "does not apply on cuda"
+    },
+    {
+      "check": "metal_tensor_runtime",
+      "arch": null,
+      "status": "SKIP",
+      "summary": "does not apply on cuda"
+    },
+    {
+      "check": "poison_probe",
+      "arch": null,
+      "status": "SKIP",
+      "summary": "no poison-probe expectation recorded for this profile"
+    },
+    {
+      "check": "text_baseline",
+      "arch": "nemotron_h_omni",
+      "status": "PASS",
+      "summary": "in-image text prefix = 19 (calibrated on 256x144+512x288); text-only = 20; text-only is +1 and MUST NOT be used (B8)",
+      "actual": 19
+    },
+    {
+      "check": "token_ladder",
+      "arch": "nemotron_h_omni",
+      "status": "PASS",
+      "summary": "5/5 geometries within +/-2",
+      "expected": {
+        "256x144": 266,
+        "512x288": 266,
+        "1024x576": 578,
+        "2048x1152": 2306,
+        "3072x1728": 3270
+      },
+      "actual": {
+        "256x144": 266,
+        "512x288": 266,
+        "1024x576": 578,
+        "2048x1152": 2306,
+        "3072x1728": 3270
+      },
+      "rows": [
+        {
+          "size": "256x144",
+          "expected": 266,
+          "actual": 266,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.01
+        },
+        {
+          "size": "512x288",
+          "expected": 266,
+          "actual": 266,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.01
+        },
+        {
+          "size": "1024x576",
+          "expected": 578,
+          "actual": 578,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.01
+        },
+        {
+          "size": "2048x1152",
+          "expected": 2306,
+          "actual": 2306,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.01
+        },
+        {
+          "size": "3072x1728",
+          "expected": 3270,
+          "actual": 3270,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.01
+        }
+      ],
+      "shape": "scaling",
+      "expected_shape": "dynamic"
+    },
+    {
+      "check": "payload_proof",
+      "arch": "nemotron_h_omni",
+      "status": "PASS",
+      "summary": "budget logged as custom on max/min (min 256*32^2=262144, max 3328*32^2=3407872)",
+      "expected": "min 256*32^2=262144, max 3328*32^2=3407872",
+      "actual": {
+        "min": "262144 (custom value)",
+        "max": "3407872 (custom value)"
+      }
+    },
+    {
+      "check": "aspect_ladder",
+      "arch": "nemotron_h_omni",
+      "status": "SKIP",
+      "summary": "no aspect-ladder expectation recorded"
+    },
+    {
+      "check": "think_format",
+      "arch": "nemotron_h_omni",
+      "status": "PASS",
+      "summary": "valid JSON after thinking (486 tokens)",
+      "expected": "non-empty JSON response with think:true + format:json",
+      "actual": {
+        "response_chars": 256,
+        "thinking_chars": 1623,
+        "eval_count": 486
+      }
+    },
+    {
+      "check": "pinned_image_token_budget",
+      "arch": "nemotron_h_omni",
+      "status": "PASS",
+      "summary": "pinned 3328 -> 3270 (ceiling 3328, +2 markers)",
+      "expected": 3270,
+      "actual": 3270,
+      "arms": [
+        {
+          "arm": "pinned",
+          "pin": 3328,
+          "expected": 3270,
+          "actual": 3270,
+          "ok": true
+        },
+        {
+          "arm": "ceiling_invariant",
+          "expected": "grid <= 3328",
+          "actual": 3268,
+          "ok": true
+        },
+        {
+          "arm": "unpinned_control",
+          "expected": 2306,
+          "actual": 2306,
+          "ok": true
+        }
+      ]
+    },
+    {
+      "check": "text_baseline",
+      "arch": "gemma4",
+      "status": "PASS",
+      "summary": "in-image text prefix = 19 (calibrated on 256x144+512x288); text-only = 19",
+      "actual": 19
+    },
+    {
+      "check": "token_ladder",
+      "arch": "gemma4",
+      "status": "PASS",
+      "summary": "5/5 geometries within +/-2",
+      "expected": {
+        "256x144": 1102,
+        "512x288": 1102,
+        "1024x576": 1102,
+        "2048x1152": 1102,
+        "3072x1728": 1102
+      },
+      "actual": {
+        "256x144": 1102,
+        "512x288": 1102,
+        "1024x576": 1102,
+        "2048x1152": 1102,
+        "3072x1728": 1102
+      },
+      "rows": [
+        {
+          "size": "256x144",
+          "expected": 1102,
+          "actual": 1102,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.0
+        },
+        {
+          "size": "512x288",
+          "expected": 1102,
+          "actual": 1102,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.01
+        },
+        {
+          "size": "1024x576",
+          "expected": 1102,
+          "actual": 1102,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.01
+        },
+        {
+          "size": "2048x1152",
+          "expected": 1102,
+          "actual": 1102,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.01
+        },
+        {
+          "size": "3072x1728",
+          "expected": 1102,
+          "actual": 1102,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.01
+        }
+      ],
+      "shape": "flat",
+      "expected_shape": "flat"
+    },
+    {
+      "check": "payload_proof",
+      "arch": "gemma4",
+      "status": "PASS",
+      "summary": "budget logged as custom on max/min (min 70*48^2=161280, max 1120*48^2=2580480)",
+      "expected": "min 70*48^2=161280, max 1120*48^2=2580480",
+      "actual": {
+        "min": "161280 (custom value)",
+        "max": "2580480 (custom value)"
+      }
+    },
+    {
+      "check": "aspect_ladder",
+      "arch": "gemma4",
+      "status": "SKIP",
+      "summary": "no aspect-ladder expectation recorded"
+    },
+    {
+      "check": "think_format",
+      "arch": "gemma4",
+      "status": "PASS",
+      "summary": "valid JSON after thinking (162 tokens)",
+      "expected": "non-empty JSON response with think:true + format:json",
+      "actual": {
+        "response_chars": 231,
+        "thinking_chars": 456,
+        "eval_count": 162
+      }
+    },
+    {
+      "check": "pinned_image_token_budget",
+      "arch": "gemma4",
+      "status": "PASS",
+      "summary": "pinned 560 -> 529 (ceiling 1120, +2 markers)",
+      "expected": 529,
+      "actual": 529,
+      "arms": [
+        {
+          "arm": "pinned",
+          "pin": 560,
+          "expected": 529,
+          "actual": 529,
+          "ok": true
+        },
+        {
+          "arm": "ceiling_invariant",
+          "expected": "grid <= 1120",
+          "actual": 527,
+          "ok": true
+        }
+      ]
+    },
+    {
+      "check": "text_baseline",
+      "arch": "qwen35",
+      "status": "PASS",
+      "summary": "in-image text prefix = 13 (calibrated on 256x144+512x288); text-only = 13",
+      "actual": 13
+    },
+    {
+      "check": "token_ladder",
+      "arch": "qwen35",
+      "status": "PASS",
+      "summary": "5/5 geometries within +/-2",
+      "expected": {
+        "256x144": 1034,
+        "512x288": 1034,
+        "1024x576": 1034,
+        "2048x1152": 2306,
+        "3072x1728": 4082
+      },
+      "actual": {
+        "256x144": 1034,
+        "512x288": 1034,
+        "1024x576": 1034,
+        "2048x1152": 2306,
+        "3072x1728": 4082
+      },
+      "rows": [
+        {
+          "size": "256x144",
+          "expected": 1034,
+          "actual": 1034,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.0
+        },
+        {
+          "size": "512x288",
+          "expected": 1034,
+          "actual": 1034,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.0
+        },
+        {
+          "size": "1024x576",
+          "expected": 1034,
+          "actual": 1034,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.0
+        },
+        {
+          "size": "2048x1152",
+          "expected": 2306,
+          "actual": 2306,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.01
+        },
+        {
+          "size": "3072x1728",
+          "expected": 4082,
+          "actual": 4082,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.01
+        }
+      ],
+      "shape": "scaling",
+      "expected_shape": "dynamic"
+    },
+    {
+      "check": "payload_proof",
+      "arch": "qwen35",
+      "status": "PASS",
+      "summary": "budget logged as custom on min (min 1024*32^2=1048576, max 4096*32^2=4194304)",
+      "expected": "min 1024*32^2=1048576, max 4096*32^2=4194304",
+      "actual": {
+        "min": "1048576 (custom value)",
+        "max": "4194304"
+      }
+    },
+    {
+      "check": "aspect_ladder",
+      "arch": "qwen35",
+      "status": "SKIP",
+      "summary": "no aspect-ladder expectation recorded"
+    },
+    {
+      "check": "think_format",
+      "arch": "qwen35",
+      "status": "PASS",
+      "summary": "valid JSON after thinking (324 tokens)",
+      "expected": "non-empty JSON response with think:true + format:json",
+      "actual": {
+        "response_chars": 414,
+        "thinking_chars": 923,
+        "eval_count": 324
+      }
+    },
+    {
+      "check": "pinned_image_token_budget",
+      "arch": "qwen35",
+      "status": "SKIP",
+      "summary": "pinned budget does not apply to this arch: image_min_tokens/image_max_tokens are arch-gated to gemma4 and nemotron_h_omni; qwen35's visionServerArgs branch ignores them."
+    },
+    {
+      "check": "metal_tensor_payload",
+      "arch": null,
+      "status": "SKIP",
+      "summary": "does not apply on cuda"
+    },
+    {
+      "check": "mlx_payload_pin",
+      "arch": null,
+      "status": "PASS",
+      "summary": "MLX build a59cc23 matches the measured payload (0.32.3-53-ga59cc23), from the shipped libmlx.so \u2014 no MLX request ran in this window, so binary/payload skew is not covered",
+      "expected": "a59cc23190f1e26fabcba8693607bb4c34b2ac2b",
+      "actual": "0.32.3-53-ga59cc23"
+    },
+    {
+      "check": "endpoint_exclusive",
+      "arch": null,
+      "status": "PASS",
+      "summary": "no contention detected (worst queue wait 0.0s)",
+      "expected": "queue wait <= 10.0s",
+      "actual": "0.0s",
+      "models_loaded": [
+        "qwen3.8:27b-q4_K_M"
+      ]
+    }
+  ],
+  "summary": {
+    "PASS": 20,
+    "SKIP": 9
+  }
+}
\ No newline at end of file
diff --git a/docs/maxusai/vision-suite/preflight/runs/preflight-cuda-0403-prod-g75c6d35-contended.json b/docs/maxusai/vision-suite/preflight/runs/preflight-cuda-0403-prod-g75c6d35-contended.json
new file mode 100644
index 00000000000..7856dec93a1
--- /dev/null
+++ b/docs/maxusai/vision-suite/preflight/runs/preflight-cuda-0403-prod-g75c6d35-contended.json
@@ -0,0 +1,476 @@
+{
+  "meta": {
+    "host": "http://127.0.0.1:11497",
+    "platform": "cuda",
+    "image_tag": "maxusai/ollama:main-75c6d350",
+    "started_utc": "20261013T215652",
+    "expectations_file": "expectations.toml",
+    "schema_version": 1,
+    "version": "0.40.2-dynres.1-17-g75c6d35",
+    "profile": "cuda-0-40-2",
+    "patchset": [
+      "001",
+      "002",
+      "004",
+      "005",
+      "801",
+      "802",
+      "805",
+      "903",
+      "908"
+    ],
+    "container": "ollama-0.40.2-dynres.1-17-g75c6d35",
+    "llama_cpp_build": "631109b34",
+    "mlx_build": "0.32.3-53-ga59cc23"
+  },
+  "results": [
+    {
+      "check": "version",
+      "arch": null,
+      "status": "PASS",
+      "summary": "0.40.2-dynres.1-17-g75c6d35 matches profile cuda-0-40-2",
+      "expected": "^0\\.40\\.2-dynres(\\.\\d+)?-\\d+-g[0-9a-f]{7,40}$",
+      "actual": "0.40.2-dynres.1-17-g75c6d35"
+    },
+    {
+      "check": "payload_pin",
+      "arch": null,
+      "status": "PASS",
+      "summary": "llama.cpp build 631109b34 matches the measured payload (docker exec)",
+      "expected": "631109b34",
+      "actual": "631109b34"
+    },
+    {
+      "check": "toolchain_pin",
+      "arch": null,
+      "status": "SKIP",
+      "summary": "profile records no toolchain_build to assert against",
+      "diagnosis": "Add toolchain_build (e.g. \"rocm-7.2.4\" or \"cuda-13.0\") so a ROCm/CUDA bump under an unchanged payload fails loudly instead of inheriting these expectations. Platforms with no such toolchain (metal, cpu) leave it unset and keep this skip; the MLX library is pinned separately by mlx_build."
+    },
+    {
+      "check": "image_tag",
+      "arch": null,
+      "status": "PASS",
+      "summary": "ollama-0.40.2-dynres.1-17-g75c6d35 runs maxusai/ollama:main-75c6d350",
+      "expected": "maxusai/ollama:main-75c6d350",
+      "actual": "maxusai/ollama:main-75c6d350"
+    },
+    {
+      "check": "go_patch_marker",
+      "arch": null,
+      "status": "PASS",
+      "summary": "--image-max-tokens present in binary (1)",
+      "expected": 1,
+      "actual": 1
+    },
+    {
+      "check": "metal_tensor_host",
+      "arch": null,
+      "status": "SKIP",
+      "summary": "does not apply on cuda"
+    },
+    {
+      "check": "metal_tensor_runtime",
+      "arch": null,
+      "status": "SKIP",
+      "summary": "does not apply on cuda"
+    },
+    {
+      "check": "poison_probe",
+      "arch": null,
+      "status": "SKIP",
+      "summary": "no poison-probe expectation recorded for this profile"
+    },
+    {
+      "check": "text_baseline",
+      "arch": "nemotron_h_omni",
+      "status": "PASS",
+      "summary": "in-image text prefix = 19 (calibrated on 256x144+512x288); text-only = 20; text-only is +1 and MUST NOT be used (B8)",
+      "actual": 19
+    },
+    {
+      "check": "token_ladder",
+      "arch": "nemotron_h_omni",
+      "status": "PASS",
+      "summary": "5/5 geometries within +/-2",
+      "expected": {
+        "256x144": 266,
+        "512x288": 266,
+        "1024x576": 578,
+        "2048x1152": 2306,
+        "3072x1728": 3270
+      },
+      "actual": {
+        "256x144": 266,
+        "512x288": 266,
+        "1024x576": 578,
+        "2048x1152": 2306,
+        "3072x1728": 3270
+      },
+      "rows": [
+        {
+          "size": "256x144",
+          "expected": 266,
+          "actual": 266,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.01
+        },
+        {
+          "size": "512x288",
+          "expected": 266,
+          "actual": 266,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.01
+        },
+        {
+          "size": "1024x576",
+          "expected": 578,
+          "actual": 578,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.01
+        },
+        {
+          "size": "2048x1152",
+          "expected": 2306,
+          "actual": 2306,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.01
+        },
+        {
+          "size": "3072x1728",
+          "expected": 3270,
+          "actual": 3270,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.01
+        }
+      ],
+      "shape": "scaling",
+      "expected_shape": "dynamic"
+    },
+    {
+      "check": "payload_proof",
+      "arch": "nemotron_h_omni",
+      "status": "PASS",
+      "summary": "budget logged as custom on max/min (min 256*32^2=262144, max 3328*32^2=3407872)",
+      "expected": "min 256*32^2=262144, max 3328*32^2=3407872",
+      "actual": {
+        "min": "262144 (custom value)",
+        "max": "3407872 (custom value)"
+      }
+    },
+    {
+      "check": "aspect_ladder",
+      "arch": "nemotron_h_omni",
+      "status": "SKIP",
+      "summary": "no aspect-ladder expectation recorded"
+    },
+    {
+      "check": "think_format",
+      "arch": "nemotron_h_omni",
+      "status": "PASS",
+      "summary": "valid JSON after thinking (486 tokens)",
+      "expected": "non-empty JSON response with think:true + format:json",
+      "actual": {
+        "response_chars": 256,
+        "thinking_chars": 1623,
+        "eval_count": 486
+      }
+    },
+    {
+      "check": "pinned_image_token_budget",
+      "arch": "nemotron_h_omni",
+      "status": "PASS",
+      "summary": "pinned 3328 -> 3270 (ceiling 3328, +2 markers)",
+      "expected": 3270,
+      "actual": 3270,
+      "arms": [
+        {
+          "arm": "pinned",
+          "pin": 3328,
+          "expected": 3270,
+          "actual": 3270,
+          "ok": true
+        },
+        {
+          "arm": "ceiling_invariant",
+          "expected": "grid <= 3328",
+          "actual": 3268,
+          "ok": true
+        },
+        {
+          "arm": "unpinned_control",
+          "expected": 2306,
+          "actual": 2306,
+          "ok": true
+        }
+      ]
+    },
+    {
+      "check": "text_baseline",
+      "arch": "gemma4",
+      "status": "PASS",
+      "summary": "in-image text prefix = -554 (calibrated on 256x144+512x288); text-only = 19; text-only is +573 and MUST NOT be used (B8)",
+      "actual": -554
+    },
+    {
+      "check": "token_ladder",
+      "arch": "gemma4",
+      "status": "FAIL",
+      "summary": "0/5 geometries within +/-2 (shape: flat, expected flat)",
+      "expected": {
+        "256x144": 1102,
+        "512x288": 1102,
+        "1024x576": 1102,
+        "2048x1152": 1102,
+        "3072x1728": 1102
+      },
+      "actual": {
+        "256x144": 1675,
+        "512x288": 1675,
+        "1024x576": 1675,
+        "2048x1152": 1675,
+        "3072x1728": 1675
+      },
+      "diagnosis": "Values moved but the shape is right: most likely a budget or preprocessing change. Re-measure and update expectations.toml if the new numbers are intended.",
+      "rows": [
+        {
+          "size": "256x144",
+          "expected": 1102,
+          "actual": 1675,
+          "delta": 573,
+          "ok": false,
+          "queue_wait_s": 0.01
+        },
+        {
+          "size": "512x288",
+          "expected": 1102,
+          "actual": 1675,
+          "delta": 573,
+          "ok": false,
+          "queue_wait_s": 0.01
+        },
+        {
+          "size": "1024x576",
+          "expected": 1102,
+          "actual": 1675,
+          "delta": 573,
+          "ok": false,
+          "queue_wait_s": 0.01
+        },
+        {
+          "size": "2048x1152",
+          "expected": 1102,
+          "actual": 1675,
+          "delta": 573,
+          "ok": false,
+          "queue_wait_s": 0.01
+        },
+        {
+          "size": "3072x1728",
+          "expected": 1102,
+          "actual": 1675,
+          "delta": 573,
+          "ok": false,
+          "queue_wait_s": 0.01
+        }
+      ],
+      "shape": "flat",
+      "expected_shape": "flat"
+    },
+    {
+      "check": "payload_proof",
+      "arch": "gemma4",
+      "status": "PASS",
+      "summary": "budget logged as custom on max/min (min 70*48^2=161280, max 1120*48^2=2580480)",
+      "expected": "min 70*48^2=161280, max 1120*48^2=2580480",
+      "actual": {
+        "min": "161280 (custom value)",
+        "max": "2580480 (custom value)"
+      }
+    },
+    {
+      "check": "aspect_ladder",
+      "arch": "gemma4",
+      "status": "SKIP",
+      "summary": "no aspect-ladder expectation recorded"
+    },
+    {
+      "check": "think_format",
+      "arch": "gemma4",
+      "status": "PASS",
+      "summary": "valid JSON after thinking (162 tokens)",
+      "expected": "non-empty JSON response with think:true + format:json",
+      "actual": {
+        "response_chars": 231,
+        "thinking_chars": 456,
+        "eval_count": 162
+      }
+    },
+    {
+      "check": "pinned_image_token_budget",
+      "arch": "gemma4",
+      "status": "FAIL",
+      "summary": "pinned 560 -> 1102 (ceiling 1120, +2 markers)",
+      "expected": 529,
+      "actual": 1102,
+      "diagnosis": "Pinned cost moved (529 -> 1102) but stays under the ceiling, so this is a behaviour change, not the 005 defect. Re-measure and update expectations.toml if intended.",
+      "arms": [
+        {
+          "arm": "pinned",
+          "pin": 560,
+          "expected": 529,
+          "actual": 1102,
+          "ok": false
+        },
+        {
+          "arm": "ceiling_invariant",
+          "expected": "grid <= 1120",
+          "actual": 1100,
+          "ok": true
+        }
+      ]
+    },
+    {
+      "check": "text_baseline",
+      "arch": "qwen35",
+      "status": "PASS",
+      "summary": "in-image text prefix = 13 (calibrated on 256x144+512x288); text-only = 13",
+      "actual": 13
+    },
+    {
+      "check": "token_ladder",
+      "arch": "qwen35",
+      "status": "PASS",
+      "summary": "5/5 geometries within +/-2",
+      "expected": {
+        "256x144": 1034,
+        "512x288": 1034,
+        "1024x576": 1034,
+        "2048x1152": 2306,
+        "3072x1728": 4082
+      },
+      "actual": {
+        "256x144": 1034,
+        "512x288": 1034,
+        "1024x576": 1034,
+        "2048x1152": 2306,
+        "3072x1728": 4082
+      },
+      "rows": [
+        {
+          "size": "256x144",
+          "expected": 1034,
+          "actual": 1034,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.0
+        },
+        {
+          "size": "512x288",
+          "expected": 1034,
+          "actual": 1034,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.01
+        },
+        {
+          "size": "1024x576",
+          "expected": 1034,
+          "actual": 1034,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.0
+        },
+        {
+          "size": "2048x1152",
+          "expected": 2306,
+          "actual": 2306,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.01
+        },
+        {
+          "size": "3072x1728",
+          "expected": 4082,
+          "actual": 4082,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.01
+        }
+      ],
+      "shape": "scaling",
+      "expected_shape": "dynamic"
+    },
+    {
+      "check": "payload_proof",
+      "arch": "qwen35",
+      "status": "PASS",
+      "summary": "budget logged as custom on min (min 1024*32^2=1048576, max 4096*32^2=4194304)",
+      "expected": "min 1024*32^2=1048576, max 4096*32^2=4194304",
+      "actual": {
+        "min": "1048576 (custom value)",
+        "max": "4194304"
+      }
+    },
+    {
+      "check": "aspect_ladder",
+      "arch": "qwen35",
+      "status": "SKIP",
+      "summary": "no aspect-ladder expectation recorded"
+    },
+    {
+      "check": "think_format",
+      "arch": "qwen35",
+      "status": "PASS",
+      "summary": "valid JSON after thinking (324 tokens)",
+      "expected": "non-empty JSON response with think:true + format:json",
+      "actual": {
+        "response_chars": 414,
+        "thinking_chars": 923,
+        "eval_count": 324
+      }
+    },
+    {
+      "check": "pinned_image_token_budget",
+      "arch": "qwen35",
+      "status": "SKIP",
+      "summary": "pinned budget does not apply to this arch: image_min_tokens/image_max_tokens are arch-gated to gemma4 and nemotron_h_omni; qwen35's visionServerArgs branch ignores them."
+    },
+    {
+      "check": "metal_tensor_payload",
+      "arch": null,
+      "status": "SKIP",
+      "summary": "does not apply on cuda"
+    },
+    {
+      "check": "mlx_payload_pin",
+      "arch": null,
+      "status": "PASS",
+      "summary": "MLX build a59cc23 matches the measured payload (0.32.3-53-ga59cc23), from the shipped libmlx.so \u2014 no MLX request ran in this window, so binary/payload skew is not covered",
+      "expected": "a59cc23190f1e26fabcba8693607bb4c34b2ac2b",
+      "actual": "0.32.3-53-ga59cc23"
+    },
+    {
+      "check": "endpoint_exclusive",
+      "arch": null,
+      "status": "PASS",
+      "summary": "no contention detected (worst queue wait 0.0s)",
+      "expected": "queue wait <= 10.0s",
+      "actual": "0.0s",
+      "models_loaded": [
+        "qwen3.8:27b-q4_K_M",
+        "gemma4:31b-it-q4_K_M"
+      ]
+    }
+  ],
+  "summary": {
+    "PASS": 18,
+    "SKIP": 9,
+    "FAIL": 2
+  }
+}
\ No newline at end of file
diff --git a/docs/maxusai/vision-suite/preflight/runs/preflight-cuda-0403-prod-g75c6d35.host-profile.json b/docs/maxusai/vision-suite/preflight/runs/preflight-cuda-0403-prod-g75c6d35.host-profile.json
new file mode 100644
index 00000000000..8d37701bc59
--- /dev/null
+++ b/docs/maxusai/vision-suite/preflight/runs/preflight-cuda-0403-prod-g75c6d35.host-profile.json
@@ -0,0 +1,145 @@
+{
+  "schema": "host-profile/1",
+  "collected_at": "2026-10-13T21:56:52Z",
+  "label": "ai-server/mlx-cuda",
+  "label_host": "ai-server",
+  "label_surface": "mlx-cuda",
+  "machine": {
+    "vendor": "ASUSTeK COMPUTER INC.",
+    "model": "WS-C621E-SAGE Series",
+    "model_id": null,
+    "virtualized": false,
+    "hypervisor": null
+  },
+  "cloud": null,
+  "os": {
+    "family": "linux",
+    "name": "Ubuntu",
+    "version": "24.04",
+    "pretty_name": "Ubuntu 24.04.5 LTS",
+    "kernel": "6.8.0-134-generic",
+    "arch": "x86_64"
+  },
+  "cpu": {
+    "vendor": "intel",
+    "model": "Intel(R) Xeon(R) Silver 4110 CPU @ 2.10GHz",
+    "logical_cpus": 32,
+    "sockets": 2,
+    "cores": 16,
+    "threads_per_core": 2,
+    "performance_cores": null,
+    "efficiency_cores": null,
+    "numa_nodes": 2,
+    "l3_cache_mib": 22.0,
+    "max_mhz": 3000.0,
+    "flags": {
+      "avx2": true,
+      "avx512f": true,
+      "avx512_vnni": false,
+      "avx512_bf16": false,
+      "avx512_fp16": false,
+      "amx_tile": false,
+      "amx_bf16": false,
+      "amx_int8": false
+    }
+  },
+  "memory_total_gib": 754.5,
+  "gpu_count": 2,
+  "gpus": [
+    {
+      "vendor": "nvidia",
+      "name": "NVIDIA RTX PRO 6000 Blackwell Workstation Edition",
+      "pci_name": "Device 2bb1",
+      "pci_bus_id": "0000:3b:00.0",
+      "pci_vendor_id": "0x10de",
+      "pci_device_id": "0x2bb1",
+      "driver": "nvidia",
+      "memory_total_mib": 97887,
+      "memory_gtt_mib": null,
+      "unified_memory": null,
+      "compute_capability": "12.0",
+      "gfx_target": null,
+      "gpu_cores": null,
+      "metal_family": null,
+      "vbios": "98.02.81.00.07",
+      "power_limit_w": 600.0,
+      "power_max_limit_w": 600.0,
+      "max_core_clock_mhz": 3090,
+      "max_memory_clock_mhz": 14001,
+      "ecc_enabled": false,
+      "mig_enabled": null,
+      "persistence_enabled": false,
+      "pcie_link_gen": 3,
+      "pcie_link_width": 16,
+      "pcie_device_max_gen": 5,
+      "pcie_device_max_width": 16,
+      "pcie_upstream_max_gen": 3,
+      "pcie_upstream_max_width": 16,
+      "pcie_gen_now": 1,
+      "pcie_width_now": 16,
+      "numa_node": 0,
+      "local_cpus": "0-7,16-23",
+      "index": 0
+    },
+    {
+      "vendor": "nvidia",
+      "name": "NVIDIA GeForce RTX 2080 Ti",
+      "pci_name": "TU102 [GeForce RTX 2080 Ti]",
+      "pci_bus_id": "0000:86:00.0",
+      "pci_vendor_id": "0x10de",
+      "pci_device_id": "0x1e04",
+      "driver": "nvidia",
+      "memory_total_mib": 11264,
+      "memory_gtt_mib": null,
+      "unified_memory": null,
+      "compute_capability": "7.5",
+      "gfx_target": null,
+      "gpu_cores": null,
+      "metal_family": null,
+      "vbios": "90.02.17.00.B2",
+      "power_limit_w": 250.0,
+      "power_max_limit_w": 280.0,
+      "max_core_clock_mhz": 2100,
+      "max_memory_clock_mhz": 7000,
+      "ecc_enabled": null,
+      "mig_enabled": null,
+      "persistence_enabled": false,
+      "pcie_link_gen": 3,
+      "pcie_link_width": 16,
+      "pcie_device_max_gen": 3,
+      "pcie_device_max_width": 16,
+      "pcie_upstream_max_gen": 3,
+      "pcie_upstream_max_width": 16,
+      "pcie_gen_now": 1,
+      "pcie_width_now": 16,
+      "numa_node": 1,
+      "local_cpus": "8-15,24-31",
+      "index": 1
+    }
+  ],
+  "filesystems": [
+    {
+      "role": "models",
+      "fstype": "btrfs",
+      "size_gib": 7452.0,
+      "raid_level": null,
+      "raid_devices": null
+    }
+  ],
+  "software": {
+    "nvidia_driver_version": "580.126.18",
+    "nvidia_kernel_module": "open",
+    "cuda_driver_version": "13.0",
+    "cuda_toolkit_version": "12.1.105",
+    "cudnn_version": "9.13.1.26",
+    "nccl_version": "2.28.3",
+    "rocm_version": null,
+    "amdgpu_driver_version": null,
+    "docker_version": "28.4.0",
+    "nvidia_container_toolkit_version": "1.20.1",
+    "dcgm_version": "3.3.9",
+    "ollama_version": "0.40.2-dynres.1-17-g75c6d35",
+    "go_version": null,
+    "python_version": "3.12.3"
+  }
+}
diff --git a/docs/maxusai/vision-suite/preflight/runs/preflight-cuda-0403-prod-g75c6d35.json b/docs/maxusai/vision-suite/preflight/runs/preflight-cuda-0403-prod-g75c6d35.json
new file mode 100644
index 00000000000..7528e4dac4c
--- /dev/null
+++ b/docs/maxusai/vision-suite/preflight/runs/preflight-cuda-0403-prod-g75c6d35.json
@@ -0,0 +1,474 @@
+{
+  "meta": {
+    "host": "http://127.0.0.1:11497",
+    "platform": "cuda",
+    "image_tag": "maxusai/ollama:main-75c6d350",
+    "started_utc": "20261013T224619",
+    "expectations_file": "expectations.toml",
+    "schema_version": 1,
+    "version": "0.40.2-dynres.1-17-g75c6d35",
+    "profile": "cuda-0-40-2",
+    "patchset": [
+      "001",
+      "002",
+      "004",
+      "005",
+      "801",
+      "802",
+      "805",
+      "903",
+      "908"
+    ],
+    "container": "ollama-0.40.2-dynres.1-17-g75c6d35",
+    "llama_cpp_build": "631109b34",
+    "mlx_build": "0.32.3-53-ga59cc23"
+  },
+  "results": [
+    {
+      "check": "version",
+      "arch": null,
+      "status": "PASS",
+      "summary": "0.40.2-dynres.1-17-g75c6d35 matches profile cuda-0-40-2",
+      "expected": "^0\\.40\\.2-dynres(\\.\\d+)?-\\d+-g[0-9a-f]{7,40}$",
+      "actual": "0.40.2-dynres.1-17-g75c6d35"
+    },
+    {
+      "check": "payload_pin",
+      "arch": null,
+      "status": "PASS",
+      "summary": "llama.cpp build 631109b34 matches the measured payload (docker exec)",
+      "expected": "631109b34",
+      "actual": "631109b34"
+    },
+    {
+      "check": "toolchain_pin",
+      "arch": null,
+      "status": "SKIP",
+      "summary": "profile records no toolchain_build to assert against",
+      "diagnosis": "Add toolchain_build (e.g. \"rocm-7.2.4\" or \"cuda-13.0\") so a ROCm/CUDA bump under an unchanged payload fails loudly instead of inheriting these expectations. Platforms with no such toolchain (metal, cpu) leave it unset and keep this skip; the MLX library is pinned separately by mlx_build."
+    },
+    {
+      "check": "image_tag",
+      "arch": null,
+      "status": "PASS",
+      "summary": "ollama-0.40.2-dynres.1-17-g75c6d35 runs maxusai/ollama:main-75c6d350",
+      "expected": "maxusai/ollama:main-75c6d350",
+      "actual": "maxusai/ollama:main-75c6d350"
+    },
+    {
+      "check": "go_patch_marker",
+      "arch": null,
+      "status": "PASS",
+      "summary": "--image-max-tokens present in binary (1)",
+      "expected": 1,
+      "actual": 1
+    },
+    {
+      "check": "metal_tensor_host",
+      "arch": null,
+      "status": "SKIP",
+      "summary": "does not apply on cuda"
+    },
+    {
+      "check": "metal_tensor_runtime",
+      "arch": null,
+      "status": "SKIP",
+      "summary": "does not apply on cuda"
+    },
+    {
+      "check": "poison_probe",
+      "arch": null,
+      "status": "SKIP",
+      "summary": "no poison-probe expectation recorded for this profile"
+    },
+    {
+      "check": "text_baseline",
+      "arch": "nemotron_h_omni",
+      "status": "PASS",
+      "summary": "in-image text prefix = 19 (calibrated on 256x144+512x288); text-only = 20; text-only is +1 and MUST NOT be used (B8)",
+      "actual": 19
+    },
+    {
+      "check": "token_ladder",
+      "arch": "nemotron_h_omni",
+      "status": "PASS",
+      "summary": "5/5 geometries within +/-2",
+      "expected": {
+        "256x144": 266,
+        "512x288": 266,
+        "1024x576": 578,
+        "2048x1152": 2306,
+        "3072x1728": 3270
+      },
+      "actual": {
+        "256x144": 266,
+        "512x288": 266,
+        "1024x576": 578,
+        "2048x1152": 2306,
+        "3072x1728": 3270
+      },
+      "rows": [
+        {
+          "size": "256x144",
+          "expected": 266,
+          "actual": 266,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.0
+        },
+        {
+          "size": "512x288",
+          "expected": 266,
+          "actual": 266,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.0
+        },
+        {
+          "size": "1024x576",
+          "expected": 578,
+          "actual": 578,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.01
+        },
+        {
+          "size": "2048x1152",
+          "expected": 2306,
+          "actual": 2306,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.01
+        },
+        {
+          "size": "3072x1728",
+          "expected": 3270,
+          "actual": 3270,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.01
+        }
+      ],
+      "shape": "scaling",
+      "expected_shape": "dynamic"
+    },
+    {
+      "check": "payload_proof",
+      "arch": "nemotron_h_omni",
+      "status": "PASS",
+      "summary": "budget logged as custom on max/min (min 256*32^2=262144, max 3328*32^2=3407872)",
+      "expected": "min 256*32^2=262144, max 3328*32^2=3407872",
+      "actual": {
+        "min": "262144 (custom value)",
+        "max": "3407872 (custom value)"
+      }
+    },
+    {
+      "check": "aspect_ladder",
+      "arch": "nemotron_h_omni",
+      "status": "SKIP",
+      "summary": "no aspect-ladder expectation recorded"
+    },
+    {
+      "check": "think_format",
+      "arch": "nemotron_h_omni",
+      "status": "PASS",
+      "summary": "valid JSON after thinking (486 tokens)",
+      "expected": "non-empty JSON response with think:true + format:json",
+      "actual": {
+        "response_chars": 256,
+        "thinking_chars": 1623,
+        "eval_count": 486
+      }
+    },
+    {
+      "check": "pinned_image_token_budget",
+      "arch": "nemotron_h_omni",
+      "status": "PASS",
+      "summary": "pinned 3328 -> 3270 (ceiling 3328, +2 markers)",
+      "expected": 3270,
+      "actual": 3270,
+      "arms": [
+        {
+          "arm": "pinned",
+          "pin": 3328,
+          "expected": 3270,
+          "actual": 3270,
+          "ok": true
+        },
+        {
+          "arm": "ceiling_invariant",
+          "expected": "grid <= 3328",
+          "actual": 3268,
+          "ok": true
+        },
+        {
+          "arm": "unpinned_control",
+          "expected": 2306,
+          "actual": 2306,
+          "ok": true
+        }
+      ]
+    },
+    {
+      "check": "text_baseline",
+      "arch": "gemma4",
+      "status": "PASS",
+      "summary": "in-image text prefix = 19 (calibrated on 256x144+512x288); text-only = 19",
+      "actual": 19
+    },
+    {
+      "check": "token_ladder",
+      "arch": "gemma4",
+      "status": "PASS",
+      "summary": "5/5 geometries within +/-2",
+      "expected": {
+        "256x144": 1102,
+        "512x288": 1102,
+        "1024x576": 1102,
+        "2048x1152": 1102,
+        "3072x1728": 1102
+      },
+      "actual": {
+        "256x144": 1102,
+        "512x288": 1102,
+        "1024x576": 1102,
+        "2048x1152": 1102,
+        "3072x1728": 1102
+      },
+      "rows": [
+        {
+          "size": "256x144",
+          "expected": 1102,
+          "actual": 1102,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.0
+        },
+        {
+          "size": "512x288",
+          "expected": 1102,
+          "actual": 1102,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.0
+        },
+        {
+          "size": "1024x576",
+          "expected": 1102,
+          "actual": 1102,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.01
+        },
+        {
+          "size": "2048x1152",
+          "expected": 1102,
+          "actual": 1102,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.01
+        },
+        {
+          "size": "3072x1728",
+          "expected": 1102,
+          "actual": 1102,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.01
+        }
+      ],
+      "shape": "flat",
+      "expected_shape": "flat"
+    },
+    {
+      "check": "payload_proof",
+      "arch": "gemma4",
+      "status": "PASS",
+      "summary": "budget logged as custom on max/min (min 70*48^2=161280, max 1120*48^2=2580480)",
+      "expected": "min 70*48^2=161280, max 1120*48^2=2580480",
+      "actual": {
+        "min": "161280 (custom value)",
+        "max": "2580480 (custom value)"
+      }
+    },
+    {
+      "check": "aspect_ladder",
+      "arch": "gemma4",
+      "status": "SKIP",
+      "summary": "no aspect-ladder expectation recorded"
+    },
+    {
+      "check": "think_format",
+      "arch": "gemma4",
+      "status": "PASS",
+      "summary": "valid JSON after thinking (162 tokens)",
+      "expected": "non-empty JSON response with think:true + format:json",
+      "actual": {
+        "response_chars": 231,
+        "thinking_chars": 456,
+        "eval_count": 162
+      }
+    },
+    {
+      "check": "pinned_image_token_budget",
+      "arch": "gemma4",
+      "status": "PASS",
+      "summary": "pinned 560 -> 529 (ceiling 1120, +2 markers)",
+      "expected": 529,
+      "actual": 529,
+      "arms": [
+        {
+          "arm": "pinned",
+          "pin": 560,
+          "expected": 529,
+          "actual": 529,
+          "ok": true
+        },
+        {
+          "arm": "ceiling_invariant",
+          "expected": "grid <= 1120",
+          "actual": 527,
+          "ok": true
+        }
+      ]
+    },
+    {
+      "check": "text_baseline",
+      "arch": "qwen35",
+      "status": "PASS",
+      "summary": "in-image text prefix = 13 (calibrated on 256x144+512x288); text-only = 13",
+      "actual": 13
+    },
+    {
+      "check": "token_ladder",
+      "arch": "qwen35",
+      "status": "PASS",
+      "summary": "5/5 geometries within +/-2",
+      "expected": {
+        "256x144": 1034,
+        "512x288": 1034,
+        "1024x576": 1034,
+        "2048x1152": 2306,
+        "3072x1728": 4082
+      },
+      "actual": {
+        "256x144": 1034,
+        "512x288": 1034,
+        "1024x576": 1034,
+        "2048x1152": 2306,
+        "3072x1728": 4082
+      },
+      "rows": [
+        {
+          "size": "256x144",
+          "expected": 1034,
+          "actual": 1034,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.01
+        },
+        {
+          "size": "512x288",
+          "expected": 1034,
+          "actual": 1034,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.01
+        },
+        {
+          "size": "1024x576",
+          "expected": 1034,
+          "actual": 1034,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.0
+        },
+        {
+          "size": "2048x1152",
+          "expected": 2306,
+          "actual": 2306,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.01
+        },
+        {
+          "size": "3072x1728",
+          "expected": 4082,
+          "actual": 4082,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.01
+        }
+      ],
+      "shape": "scaling",
+      "expected_shape": "dynamic"
+    },
+    {
+      "check": "payload_proof",
+      "arch": "qwen35",
+      "status": "PASS",
+      "summary": "budget logged as custom on min (min 1024*32^2=1048576, max 4096*32^2=4194304)",
+      "expected": "min 1024*32^2=1048576, max 4096*32^2=4194304",
+      "actual": {
+        "min": "1048576 (custom value)",
+        "max": "4194304"
+      }
+    },
+    {
+      "check": "aspect_ladder",
+      "arch": "qwen35",
+      "status": "SKIP",
+      "summary": "no aspect-ladder expectation recorded"
+    },
+    {
+      "check": "think_format",
+      "arch": "qwen35",
+      "status": "PASS",
+      "summary": "valid JSON after thinking (324 tokens)",
+      "expected": "non-empty JSON response with think:true + format:json",
+      "actual": {
+        "response_chars": 414,
+        "thinking_chars": 923,
+        "eval_count": 324
+      }
+    },
+    {
+      "check": "pinned_image_token_budget",
+      "arch": "qwen35",
+      "status": "SKIP",
+      "summary": "pinned budget does not apply to this arch: image_min_tokens/image_max_tokens are arch-gated to gemma4 and nemotron_h_omni; qwen35's visionServerArgs branch ignores them."
+    },
+    {
+      "check": "metal_tensor_payload",
+      "arch": null,
+      "status": "SKIP",
+      "summary": "does not apply on cuda"
+    },
+    {
+      "check": "mlx_payload_pin",
+      "arch": null,
+      "status": "PASS",
+      "summary": "MLX build a59cc23 matches the measured payload (0.32.3-53-ga59cc23), from the shipped libmlx.so \u2014 no MLX request ran in this window, so binary/payload skew is not covered",
+      "expected": "a59cc23190f1e26fabcba8693607bb4c34b2ac2b",
+      "actual": "0.32.3-53-ga59cc23"
+    },
+    {
+      "check": "endpoint_exclusive",
+      "arch": null,
+      "status": "PASS",
+      "summary": "no contention detected (worst queue wait 0.0s)",
+      "expected": "queue wait <= 10.0s",
+      "actual": "0.0s",
+      "models_loaded": [
+        "qwen3.8:27b-q4_K_M",
+        "gemma4:31b-it-q4_K_M",
+        "nemotron3:33b-q8"
+      ]
+    }
+  ],
+  "summary": {
+    "PASS": 20,
+    "SKIP": 9
+  }
+}
\ No newline at end of file
diff --git a/docs/maxusai/vision-suite/preflight/runs/preflight-mlx-cuda-0403-canary-gpusall-g75c6d35.json b/docs/maxusai/vision-suite/preflight/runs/preflight-mlx-cuda-0403-canary-gpusall-g75c6d35.json
new file mode 100644
index 00000000000..7ca90811f3b
--- /dev/null
+++ b/docs/maxusai/vision-suite/preflight/runs/preflight-mlx-cuda-0403-canary-gpusall-g75c6d35.json
@@ -0,0 +1,394 @@
+{
+  "meta": {
+    "host": "http://127.0.0.1:11532",
+    "platform": "mlx-cuda",
+    "image_tag": "maxusai/ollama:main-75c6d350",
+    "started_utc": "20261013T214618",
+    "expectations_file": "expectations.toml",
+    "schema_version": 1,
+    "version": "0.40.2-dynres.1-17-g75c6d35",
+    "profile": "mlx-cuda-0-40-2",
+    "patchset": null,
+    "container": "ollama-ab-0402-canary0403-gpusall",
+    "llama_cpp_build": "631109b34",
+    "mlx_build": "0.32.3-53-ga59cc23"
+  },
+  "results": [
+    {
+      "check": "version",
+      "arch": null,
+      "status": "PASS",
+      "summary": "0.40.2-dynres.1-17-g75c6d35 matches profile mlx-cuda-0-40-2",
+      "expected": "^0\\.40\\.2-dynres(\\.\\d+)?-\\d+-g[0-9a-f]{7,40}$",
+      "actual": "0.40.2-dynres.1-17-g75c6d35"
+    },
+    {
+      "check": "payload_pin",
+      "arch": null,
+      "status": "SKIP",
+      "summary": "profile records no llama_cpp_build to assert against",
+      "diagnosis": "Add llama_cpp_build to this profile so a llama.cpp bump fails loudly instead of inheriting stale expectations. See README.md."
+    },
+    {
+      "check": "toolchain_pin",
+      "arch": null,
+      "status": "SKIP",
+      "summary": "profile records no toolchain_build to assert against",
+      "diagnosis": "Add toolchain_build (e.g. \"rocm-7.2.4\" or \"cuda-13.0\") so a ROCm/CUDA bump under an unchanged payload fails loudly instead of inheriting these expectations. Platforms with no such toolchain (metal, cpu) leave it unset and keep this skip; the MLX library is pinned separately by mlx_build."
+    },
+    {
+      "check": "image_tag",
+      "arch": null,
+      "status": "PASS",
+      "summary": "ollama-ab-0402-canary0403-gpusall runs maxusai/ollama:main-75c6d350",
+      "expected": "maxusai/ollama:main-75c6d350",
+      "actual": "maxusai/ollama:main-75c6d350"
+    },
+    {
+      "check": "go_patch_marker",
+      "arch": null,
+      "status": "PASS",
+      "summary": "--image-max-tokens present in binary (1)",
+      "expected": 1,
+      "actual": 1
+    },
+    {
+      "check": "metal_tensor_host",
+      "arch": null,
+      "status": "SKIP",
+      "summary": "does not apply on mlx-cuda"
+    },
+    {
+      "check": "metal_tensor_runtime",
+      "arch": null,
+      "status": "SKIP",
+      "summary": "does not apply on mlx-cuda"
+    },
+    {
+      "check": "poison_probe",
+      "arch": null,
+      "status": "SKIP",
+      "summary": "no poison-probe expectation recorded for this profile"
+    },
+    {
+      "check": "text_baseline",
+      "arch": "gemma4",
+      "status": "PASS",
+      "summary": "in-image text prefix = 19 (calibrated on 256x144+512x288); text-only = 19",
+      "actual": 19
+    },
+    {
+      "check": "token_ladder",
+      "arch": "gemma4",
+      "status": "PASS",
+      "summary": "5/5 geometries within +/-2",
+      "expected": {
+        "256x144": 1102,
+        "512x288": 1102,
+        "1024x576": 1102,
+        "2048x1152": 1102,
+        "3072x1728": 1102
+      },
+      "actual": {
+        "256x144": 1102,
+        "512x288": 1102,
+        "1024x576": 1102,
+        "2048x1152": 1102,
+        "3072x1728": 1102
+      },
+      "rows": [
+        {
+          "size": "256x144",
+          "expected": 1102,
+          "actual": 1102,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.01
+        },
+        {
+          "size": "512x288",
+          "expected": 1102,
+          "actual": 1102,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.01
+        },
+        {
+          "size": "1024x576",
+          "expected": 1102,
+          "actual": 1102,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.01
+        },
+        {
+          "size": "2048x1152",
+          "expected": 1102,
+          "actual": 1102,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.01
+        },
+        {
+          "size": "3072x1728",
+          "expected": 1102,
+          "actual": 1102,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.01
+        }
+      ],
+      "shape": "flat",
+      "expected_shape": "flat"
+    },
+    {
+      "check": "payload_proof",
+      "arch": "gemma4",
+      "status": "SKIP",
+      "summary": "budget/pixel values are not observable on this platform",
+      "diagnosis": "This block's budgets were established without a load log (the MLX runner emits none, on either backend) and are recorded as budgets_observed = false. Nothing here is waiting on a container; the block says how the values were established."
+    },
+    {
+      "check": "aspect_ladder",
+      "arch": "gemma4",
+      "status": "SKIP",
+      "summary": "no aspect-ladder expectation recorded"
+    },
+    {
+      "check": "think_format",
+      "arch": "gemma4",
+      "status": "SKIP",
+      "summary": "no think+format expectation recorded"
+    },
+    {
+      "check": "pinned_image_token_budget",
+      "arch": "gemma4",
+      "status": "SKIP",
+      "summary": "no pinned-budget expectation recorded for this arch",
+      "diagnosis": "Omitted deliberately: the 005 defect was found on nemotron and this arch's probe has not been measured. See README.md to add one."
+    },
+    {
+      "check": "text_baseline",
+      "arch": "qwen35moe",
+      "status": "PASS",
+      "summary": "in-image text prefix = 13 (calibrated on 256x144+512x288); text-only = 13",
+      "actual": 13
+    },
+    {
+      "check": "token_ladder",
+      "arch": "qwen35moe",
+      "status": "PASS",
+      "summary": "5/5 geometries within +/-2",
+      "expected": {
+        "256x144": 68,
+        "512x288": 146,
+        "1024x576": 578,
+        "2048x1152": 2306,
+        "3072x1728": 5186
+      },
+      "actual": {
+        "256x144": 68,
+        "512x288": 146,
+        "1024x576": 578,
+        "2048x1152": 2306,
+        "3072x1728": 5186
+      },
+      "rows": [
+        {
+          "size": "256x144",
+          "expected": 68,
+          "actual": 68,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.0
+        },
+        {
+          "size": "512x288",
+          "expected": 146,
+          "actual": 146,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.0
+        },
+        {
+          "size": "1024x576",
+          "expected": 578,
+          "actual": 578,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.01
+        },
+        {
+          "size": "2048x1152",
+          "expected": 2306,
+          "actual": 2306,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.01
+        },
+        {
+          "size": "3072x1728",
+          "expected": 5186,
+          "actual": 5186,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.01
+        }
+      ],
+      "shape": "scaling",
+      "expected_shape": "dynamic"
+    },
+    {
+      "check": "payload_proof",
+      "arch": "qwen35moe",
+      "status": "SKIP",
+      "summary": "budget/pixel values are not observable on this platform",
+      "diagnosis": "This block's budgets were established without a load log (the MLX runner emits none, on either backend) and are recorded as budgets_observed = false. Nothing here is waiting on a container; the block says how the values were established."
+    },
+    {
+      "check": "aspect_ladder",
+      "arch": "qwen35moe",
+      "status": "SKIP",
+      "summary": "no aspect-ladder expectation recorded"
+    },
+    {
+      "check": "think_format",
+      "arch": "qwen35moe",
+      "status": "SKIP",
+      "summary": "no think+format expectation recorded"
+    },
+    {
+      "check": "pinned_image_token_budget",
+      "arch": "qwen35moe",
+      "status": "SKIP",
+      "summary": "no pinned-budget expectation recorded for this arch",
+      "diagnosis": "Omitted deliberately: the 005 defect was found on nemotron and this arch's probe has not been measured. See README.md to add one."
+    },
+    {
+      "check": "text_baseline",
+      "arch": "qwen35",
+      "status": "PASS",
+      "summary": "in-image text prefix = 13 (calibrated on 256x144+512x288); text-only = 13",
+      "actual": 13
+    },
+    {
+      "check": "token_ladder",
+      "arch": "qwen35",
+      "status": "PASS",
+      "summary": "5/5 geometries within +/-2",
+      "expected": {
+        "256x144": 68,
+        "512x288": 146,
+        "1024x576": 578,
+        "2048x1152": 2306,
+        "3072x1728": 5186
+      },
+      "actual": {
+        "256x144": 68,
+        "512x288": 146,
+        "1024x576": 578,
+        "2048x1152": 2306,
+        "3072x1728": 5186
+      },
+      "rows": [
+        {
+          "size": "256x144",
+          "expected": 68,
+          "actual": 68,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.0
+        },
+        {
+          "size": "512x288",
+          "expected": 146,
+          "actual": 146,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.01
+        },
+        {
+          "size": "1024x576",
+          "expected": 578,
+          "actual": 578,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.01
+        },
+        {
+          "size": "2048x1152",
+          "expected": 2306,
+          "actual": 2306,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.01
+        },
+        {
+          "size": "3072x1728",
+          "expected": 5186,
+          "actual": 5186,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.01
+        }
+      ],
+      "shape": "scaling",
+      "expected_shape": "dynamic"
+    },
+    {
+      "check": "payload_proof",
+      "arch": "qwen35",
+      "status": "SKIP",
+      "summary": "budget/pixel values are not observable on this platform",
+      "diagnosis": "This block's budgets were established without a load log (the MLX runner emits none, on either backend) and are recorded as budgets_observed = false. Nothing here is waiting on a container; the block says how the values were established."
+    },
+    {
+      "check": "aspect_ladder",
+      "arch": "qwen35",
+      "status": "SKIP",
+      "summary": "no aspect-ladder expectation recorded"
+    },
+    {
+      "check": "think_format",
+      "arch": "qwen35",
+      "status": "SKIP",
+      "summary": "no think+format expectation recorded"
+    },
+    {
+      "check": "pinned_image_token_budget",
+      "arch": "qwen35",
+      "status": "SKIP",
+      "summary": "no pinned-budget expectation recorded for this arch",
+      "diagnosis": "Omitted deliberately: the 005 defect was found on nemotron and this arch's probe has not been measured. See README.md to add one."
+    },
+    {
+      "check": "metal_tensor_payload",
+      "arch": null,
+      "status": "SKIP",
+      "summary": "does not apply on mlx-cuda"
+    },
+    {
+      "check": "mlx_payload_pin",
+      "arch": null,
+      "status": "PASS",
+      "summary": "MLX build a59cc23 matches the measured payload (0.32.3-53-ga59cc23), from the runner's engine-init line",
+      "expected": "a59cc23190f1e26fabcba8693607bb4c34b2ac2b",
+      "actual": "0.32.3-53-ga59cc23"
+    },
+    {
+      "check": "endpoint_exclusive",
+      "arch": null,
+      "status": "PASS",
+      "summary": "no contention detected (worst queue wait 0.0s)",
+      "expected": "queue wait <= 10.0s",
+      "actual": "0.0s",
+      "models_loaded": [
+        "qwen3.8:27b-nvfp4"
+      ]
+    }
+  ],
+  "summary": {
+    "PASS": 11,
+    "SKIP": 18
+  }
+}
\ No newline at end of file
diff --git a/docs/maxusai/vision-suite/preflight/runs/preflight-mlx-cuda-0403-prod-g75c6d35.json b/docs/maxusai/vision-suite/preflight/runs/preflight-mlx-cuda-0403-prod-g75c6d35.json
new file mode 100644
index 00000000000..160efdd0cf3
--- /dev/null
+++ b/docs/maxusai/vision-suite/preflight/runs/preflight-mlx-cuda-0403-prod-g75c6d35.json
@@ -0,0 +1,395 @@
+{
+  "meta": {
+    "host": "http://127.0.0.1:11497",
+    "platform": "mlx-cuda",
+    "image_tag": "maxusai/ollama:main-75c6d350",
+    "started_utc": "20261013T221039",
+    "expectations_file": "expectations.toml",
+    "schema_version": 1,
+    "version": "0.40.2-dynres.1-17-g75c6d35",
+    "profile": "mlx-cuda-0-40-2",
+    "patchset": null,
+    "container": "ollama-0.40.2-dynres.1-17-g75c6d35",
+    "llama_cpp_build": "631109b34",
+    "mlx_build": "0.32.3-53-ga59cc23"
+  },
+  "results": [
+    {
+      "check": "version",
+      "arch": null,
+      "status": "PASS",
+      "summary": "0.40.2-dynres.1-17-g75c6d35 matches profile mlx-cuda-0-40-2",
+      "expected": "^0\\.40\\.2-dynres(\\.\\d+)?-\\d+-g[0-9a-f]{7,40}$",
+      "actual": "0.40.2-dynres.1-17-g75c6d35"
+    },
+    {
+      "check": "payload_pin",
+      "arch": null,
+      "status": "SKIP",
+      "summary": "profile records no llama_cpp_build to assert against",
+      "diagnosis": "Add llama_cpp_build to this profile so a llama.cpp bump fails loudly instead of inheriting stale expectations. See README.md."
+    },
+    {
+      "check": "toolchain_pin",
+      "arch": null,
+      "status": "SKIP",
+      "summary": "profile records no toolchain_build to assert against",
+      "diagnosis": "Add toolchain_build (e.g. \"rocm-7.2.4\" or \"cuda-13.0\") so a ROCm/CUDA bump under an unchanged payload fails loudly instead of inheriting these expectations. Platforms with no such toolchain (metal, cpu) leave it unset and keep this skip; the MLX library is pinned separately by mlx_build."
+    },
+    {
+      "check": "image_tag",
+      "arch": null,
+      "status": "PASS",
+      "summary": "ollama-0.40.2-dynres.1-17-g75c6d35 runs maxusai/ollama:main-75c6d350",
+      "expected": "maxusai/ollama:main-75c6d350",
+      "actual": "maxusai/ollama:main-75c6d350"
+    },
+    {
+      "check": "go_patch_marker",
+      "arch": null,
+      "status": "PASS",
+      "summary": "--image-max-tokens present in binary (1)",
+      "expected": 1,
+      "actual": 1
+    },
+    {
+      "check": "metal_tensor_host",
+      "arch": null,
+      "status": "SKIP",
+      "summary": "does not apply on mlx-cuda"
+    },
+    {
+      "check": "metal_tensor_runtime",
+      "arch": null,
+      "status": "SKIP",
+      "summary": "does not apply on mlx-cuda"
+    },
+    {
+      "check": "poison_probe",
+      "arch": null,
+      "status": "SKIP",
+      "summary": "no poison-probe expectation recorded for this profile"
+    },
+    {
+      "check": "text_baseline",
+      "arch": "gemma4",
+      "status": "PASS",
+      "summary": "in-image text prefix = 19 (calibrated on 256x144+512x288); text-only = 19",
+      "actual": 19
+    },
+    {
+      "check": "token_ladder",
+      "arch": "gemma4",
+      "status": "PASS",
+      "summary": "5/5 geometries within +/-2",
+      "expected": {
+        "256x144": 1102,
+        "512x288": 1102,
+        "1024x576": 1102,
+        "2048x1152": 1102,
+        "3072x1728": 1102
+      },
+      "actual": {
+        "256x144": 1102,
+        "512x288": 1102,
+        "1024x576": 1102,
+        "2048x1152": 1102,
+        "3072x1728": 1102
+      },
+      "rows": [
+        {
+          "size": "256x144",
+          "expected": 1102,
+          "actual": 1102,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.01
+        },
+        {
+          "size": "512x288",
+          "expected": 1102,
+          "actual": 1102,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.01
+        },
+        {
+          "size": "1024x576",
+          "expected": 1102,
+          "actual": 1102,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.01
+        },
+        {
+          "size": "2048x1152",
+          "expected": 1102,
+          "actual": 1102,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.01
+        },
+        {
+          "size": "3072x1728",
+          "expected": 1102,
+          "actual": 1102,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.01
+        }
+      ],
+      "shape": "flat",
+      "expected_shape": "flat"
+    },
+    {
+      "check": "payload_proof",
+      "arch": "gemma4",
+      "status": "SKIP",
+      "summary": "budget/pixel values are not observable on this platform",
+      "diagnosis": "This block's budgets were established without a load log (the MLX runner emits none, on either backend) and are recorded as budgets_observed = false. Nothing here is waiting on a container; the block says how the values were established."
+    },
+    {
+      "check": "aspect_ladder",
+      "arch": "gemma4",
+      "status": "SKIP",
+      "summary": "no aspect-ladder expectation recorded"
+    },
+    {
+      "check": "think_format",
+      "arch": "gemma4",
+      "status": "SKIP",
+      "summary": "no think+format expectation recorded"
+    },
+    {
+      "check": "pinned_image_token_budget",
+      "arch": "gemma4",
+      "status": "SKIP",
+      "summary": "no pinned-budget expectation recorded for this arch",
+      "diagnosis": "Omitted deliberately: the 005 defect was found on nemotron and this arch's probe has not been measured. See README.md to add one."
+    },
+    {
+      "check": "text_baseline",
+      "arch": "qwen35moe",
+      "status": "PASS",
+      "summary": "in-image text prefix = 13 (calibrated on 256x144+512x288); text-only = 13",
+      "actual": 13
+    },
+    {
+      "check": "token_ladder",
+      "arch": "qwen35moe",
+      "status": "PASS",
+      "summary": "5/5 geometries within +/-2",
+      "expected": {
+        "256x144": 68,
+        "512x288": 146,
+        "1024x576": 578,
+        "2048x1152": 2306,
+        "3072x1728": 5186
+      },
+      "actual": {
+        "256x144": 68,
+        "512x288": 146,
+        "1024x576": 578,
+        "2048x1152": 2306,
+        "3072x1728": 5186
+      },
+      "rows": [
+        {
+          "size": "256x144",
+          "expected": 68,
+          "actual": 68,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.01
+        },
+        {
+          "size": "512x288",
+          "expected": 146,
+          "actual": 146,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.01
+        },
+        {
+          "size": "1024x576",
+          "expected": 578,
+          "actual": 578,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.0
+        },
+        {
+          "size": "2048x1152",
+          "expected": 2306,
+          "actual": 2306,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.01
+        },
+        {
+          "size": "3072x1728",
+          "expected": 5186,
+          "actual": 5186,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.01
+        }
+      ],
+      "shape": "scaling",
+      "expected_shape": "dynamic"
+    },
+    {
+      "check": "payload_proof",
+      "arch": "qwen35moe",
+      "status": "SKIP",
+      "summary": "budget/pixel values are not observable on this platform",
+      "diagnosis": "This block's budgets were established without a load log (the MLX runner emits none, on either backend) and are recorded as budgets_observed = false. Nothing here is waiting on a container; the block says how the values were established."
+    },
+    {
+      "check": "aspect_ladder",
+      "arch": "qwen35moe",
+      "status": "SKIP",
+      "summary": "no aspect-ladder expectation recorded"
+    },
+    {
+      "check": "think_format",
+      "arch": "qwen35moe",
+      "status": "SKIP",
+      "summary": "no think+format expectation recorded"
+    },
+    {
+      "check": "pinned_image_token_budget",
+      "arch": "qwen35moe",
+      "status": "SKIP",
+      "summary": "no pinned-budget expectation recorded for this arch",
+      "diagnosis": "Omitted deliberately: the 005 defect was found on nemotron and this arch's probe has not been measured. See README.md to add one."
+    },
+    {
+      "check": "text_baseline",
+      "arch": "qwen35",
+      "status": "PASS",
+      "summary": "in-image text prefix = 13 (calibrated on 256x144+512x288); text-only = 13",
+      "actual": 13
+    },
+    {
+      "check": "token_ladder",
+      "arch": "qwen35",
+      "status": "PASS",
+      "summary": "5/5 geometries within +/-2",
+      "expected": {
+        "256x144": 68,
+        "512x288": 146,
+        "1024x576": 578,
+        "2048x1152": 2306,
+        "3072x1728": 5186
+      },
+      "actual": {
+        "256x144": 68,
+        "512x288": 146,
+        "1024x576": 578,
+        "2048x1152": 2306,
+        "3072x1728": 5186
+      },
+      "rows": [
+        {
+          "size": "256x144",
+          "expected": 68,
+          "actual": 68,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.0
+        },
+        {
+          "size": "512x288",
+          "expected": 146,
+          "actual": 146,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.0
+        },
+        {
+          "size": "1024x576",
+          "expected": 578,
+          "actual": 578,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.01
+        },
+        {
+          "size": "2048x1152",
+          "expected": 2306,
+          "actual": 2306,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.01
+        },
+        {
+          "size": "3072x1728",
+          "expected": 5186,
+          "actual": 5186,
+          "delta": 0,
+          "ok": true,
+          "queue_wait_s": 0.01
+        }
+      ],
+      "shape": "scaling",
+      "expected_shape": "dynamic"
+    },
+    {
+      "check": "payload_proof",
+      "arch": "qwen35",
+      "status": "SKIP",
+      "summary": "budget/pixel values are not observable on this platform",
+      "diagnosis": "This block's budgets were established without a load log (the MLX runner emits none, on either backend) and are recorded as budgets_observed = false. Nothing here is waiting on a container; the block says how the values were established."
+    },
+    {
+      "check": "aspect_ladder",
+      "arch": "qwen35",
+      "status": "SKIP",
+      "summary": "no aspect-ladder expectation recorded"
+    },
+    {
+      "check": "think_format",
+      "arch": "qwen35",
+      "status": "SKIP",
+      "summary": "no think+format expectation recorded"
+    },
+    {
+      "check": "pinned_image_token_budget",
+      "arch": "qwen35",
+      "status": "SKIP",
+      "summary": "no pinned-budget expectation recorded for this arch",
+      "diagnosis": "Omitted deliberately: the 005 defect was found on nemotron and this arch's probe has not been measured. See README.md to add one."
+    },
+    {
+      "check": "metal_tensor_payload",
+      "arch": null,
+      "status": "SKIP",
+      "summary": "does not apply on mlx-cuda"
+    },
+    {
+      "check": "mlx_payload_pin",
+      "arch": null,
+      "status": "PASS",
+      "summary": "MLX build a59cc23 matches the measured payload (0.32.3-53-ga59cc23), from the runner's engine-init line",
+      "expected": "a59cc23190f1e26fabcba8693607bb4c34b2ac2b",
+      "actual": "0.32.3-53-ga59cc23"
+    },
+    {
+      "check": "endpoint_exclusive",
+      "arch": null,
+      "status": "PASS",
+      "summary": "no contention detected (worst queue wait 0.0s)",
+      "expected": "queue wait <= 10.0s",
+      "actual": "0.0s",
+      "models_loaded": [
+        "qwen3.8:27b-nvfp4",
+        "qwen3.6:35b-a3b-nvfp4"
+      ]
+    }
+  ],
+  "summary": {
+    "PASS": 11,
+    "SKIP": 18
+  }
+}
\ No newline at end of file