diff --git a/devlog/_plan/260826_glm53_flash_preseed/000_plan.md b/devlog/_plan/260826_glm53_flash_preseed/000_plan.md
new file mode 100644
index 0000000000..f404ca633c
--- /dev/null
+++ b/devlog/_plan/260826_glm53_flash_preseed/000_plan.md
@@ -0,0 +1,150 @@
+# 000 — Seeding `glm-5.3-flash` ahead of the providers that will serve it
+
+## The ask, and why the codebase already agrees with it
+
+Put `glm-5.3-flash` everywhere `ox-alpha` or `glm-5.2` already lives, before the
+providers announce it.
+
+That is not a new policy. `src/providers/registry.ts:340-342` states it:
+
+> The non-Z.AI providers below are speculative on purpose: they carry 5.2 today and are
+> expected to pick 5.3 up on their usual lag. Providers whose live `/v1/models`
+> discovery is enabled self-correct on the next successful fetch; static ones need a
+> follow-up refresh.
+
+This unit is the same move one generation on. Two facts make it safe:
+
+- `rg glm-5.3-flash` returns **nothing** today, so there is no duplicate id to collide with.
+- `glm-4.7-flash` already sits beside `glm-4.7` in `ZHIPU_BIGMODEL_TEXT_MODELS`, so a
+  flash sibling next to its full model is an established shape here, not an invention.
+
+## What `glm-5.3-flash` is
+
+Z.AI's flash tier for GLM-5.3: 1M context, text-only, cheaper than the full model. Where
+a provider needs a number this unit does not have from that provider's own docs, it
+**mirrors the `glm-5.3` entry that provider already carries** rather than inventing a
+figure. That is stated per-cluster below and is the difference between seeding and guessing.
+
+## The vision decision, corrected mid-unit
+
+The first pass put `glm-5.3-flash` into every `noVisionModels` list, reasoning from
+`registry.ts:500-506`:
+
+> Verified-negative and therefore deliberately ABSENT: … zai-org/GLM-5.2, zai-org/GLM-5.3
+> … Those routes accept the request and drop the image, which is worse than declining it.
+
+**That was wrong, and the same comment says why.** It ends with "Do not add an id here on
+family resemblance" — and inheriting a text-only verdict from `glm-5.3` because of the
+shared name is that exact error, pointed the other way.
+
+`glm-5.3-flash` is a vision model. Z.AI documents it under `docs.z.ai/guides/vlm/` — the
+VLM namespace, the same one `glm-4.6v` lives in — accepting `image_url` blocks as public
+URLs or Base64, multiple images per request, plus video and file input. OpenRouter lists
+it as a multimodal reasoning model: 1M context, 128K max output, text + image + video.
+
+One incidental confirmation: this unit had already given it `ZAI_GLM_53_REASONING_EFFORTS`
+(low/high/max), which matches the three-tier ladder OpenRouter documents for it. The
+effort ladder was right for the same reason the modality was wrong — 5.3 is its family.
+
+The naming split is the tell: `-flash` is a speed and price tier; `v` is the modality
+suffix. A flash variant of a VLM is still a VLM.
+
+So it ships with `["text", "image"]` and appears in **no** `noVisionModels` list.
+
+## Ox Alpha: removed entirely
+
+The first pass excluded `glm-5.3-flash` from the Ox Alpha surfaces and left Ox Alpha in
+place. It is now removed from the tree outright — both ids (`stealth/ox-alpha`,
+`openai/ox-alpha`) and the OpenCode Zen slug that served the same stealth model
+(`x-preview-f-free`), along with `OX_ALPHA_CONTEXT_WINDOW`, the Command Code effort
+profile, the OpenRouter catalog entry, and every comment describing them.
+
+"Ox Alpha" was a stealth-window model: free, unbranded, and time-boxed by construction.
+A catalog entry for a window that has closed advertises something the provider no longer
+serves, which is the same class of defect this unit exists to avoid in the other
+direction.
+
+## Insertion points
+
+Every one of these already carries `glm-5.2`; `glm-5.3-flash` goes in beside `glm-5.3`.
+
+| # | Location | Symbol | Note |
+|---|---|---|---|
+| 1 | registry.ts:474 | `ZHIPU_BIGMODEL_TEXT_MODELS` | flash sibling precedent lives here (`glm-4.7-flash`) |
+| 2 | registry.ts:480 | `ZHIPU_BIGMODEL_THINKING_TOGGLE_MODELS` | 5.3 is present, so the toggle applies |
+| 3 | registry.ts:616 + 627 | `ALIBABA_TOKEN_PLAN_MODELS` + input modalities | `["text"]`, matching its `glm-5.3` |
+| 4 | registry.ts:639 | `ALIBABA_INTL_TOKEN_PLAN_MODELS` | + modalities at 854-ish block |
+| 5 | registry.ts:689 / 697 / 715 | Volcengine coding / agent / text-only | text-only list too |
+| 6 | registry.ts:864 | `NEURALWATT_REASONING_HISTORY_MODELS` | Neuralwatt suffixes its ids; see caution below |
+| 7 | registry.ts:944 | baseten list | |
+| 8 | registry.ts:952 | `SCALEWAY_SERVERLESS_CHAT_MODELS` | |
+| 9 | registry.ts:973 / 982 / 988 | `UMANS_MODELS` + text-only + context | prefix is `umans-` |
+| 10 | registry.ts:997 / 1012 | cline-pass | prefix is `cline-pass/` |
+| 11 | registry.ts:2201 / 2282 / 2284 | `zai` + `zhipu-bigmodel-coding` provider blocks | context window mirrors 5.3's 1M |
+| 12 | registry.ts:2428 / 2464 | alibaba plan context windows | 1M, mirroring 5.3 |
+| 13 | registry.ts:2514 / 2517 | ollama-cloud | |
+| 14 | registry.ts:2730 | cloudflare `@cf/zai-org/…` | prefix form |
+
+**Caution on 6 and 11.** Neuralwatt fans each model into `-fast` / `-short` /
+`-short-fast` variants and the parity test pins the full list with `toEqual`. Z.AI fans
+into `[1m]` aliases with three pinned `toEqual` maps. Adding a bare id to either without
+its variants leaves the tables internally inconsistent. This unit adds the **plain
+`glm-5.3-flash`** id and does not synthesize `-fast`/`-short`/`[1m]` variants: those
+suffixes encode routing behavior those providers documented per model, and inventing
+them would assert a product that may not exist.
+
+## The test surface — this is the real work
+
+`tests/provider-registry-parity.test.ts` pins exact lists with `toEqual` (17 `glm-5.3`
+hits). Every seeded list has a matching assertion that must move in the same commit, or
+the suite goes red. Also covering these tables: `volcengine-providers`,
+`cline-pass-provider`, `alibaba-intl-token-plan`, `catalog-vision-sidecar-modalities`,
+`codex-catalog`, `umans-provider`, `routing-compatibility-model-matching`.
+
+`src/generated/model-metadata.ts` is **generated** (`scripts/generate-model-metadata.ts`,
+"Do not edit by hand") and is NOT touched here. Its data comes from upstream catalogs; it
+will pick `glm-5.3-flash` up on the next regeneration once providers publish it.
+
+## Accept criteria
+
+## Where a seed actually reaches the user (measured)
+
+Seeding only matters where the static catalog is what ships. Reading each provider block:
+
+| Provider | liveModels | modelDiscovery | Seed reaches the user? |
+|---|---|---|---|
+| `zhipu-bigmodel` | false | no | **yes** — static |
+| `alibaba-token-plan` / `-intl` | false | no | **yes** — static |
+| `volcengine-coding-plan` / `-agent-plan` | false | no | **yes** — static |
+| `neuralwatt` | false | no | **yes** — static |
+| `umans` | false | no | **yes** — static |
+| `zai` | false | no | **yes** — static |
+| `ollama-cloud` | false | no | **yes** — static |
+| `zhipu-bigmodel-coding` | true | no | seed is the offline fallback |
+| `cline-pass` | true | no | seed is the offline fallback |
+| `baseten` | true | yes | overwritten on first successful fetch |
+| `scaleway` | true | yes | overwritten on first successful fetch |
+| `cloudflare-workers-ai` | true | yes | overwritten on first successful fetch |
+
+This confirms the claim quoted at the top of this document rather than assuming it: the
+static providers are exactly the ones that "need a follow-up refresh", and they are the
+majority here. The three discovery-enabled providers still get the id — it is their
+documented offline fallback, and a wrong-but-harmless entry there is replaced the moment
+a real fetch succeeds.
+
+| # | Criterion | Evidence |
+|---|---|---|
+| 1 | Every `glm-5.2` provider list also carries `glm-5.3-flash` | `rg` shows the pair per cluster |
+| 2 | No vision/image list gains it | it appears in `noVisionModels`, never `modelInputModalities` as image |
+| 3 | ox-alpha surfaces reviewed and excluded on the record | this document |
+| 4 | Parity and provider tests green | narrow `bun test` on the 7 covering files |
+| 5 | Types hold | `bun x tsc --noEmit` |
+| 6 | Landed | CI green, PR merged, local `dev` level |
+
+## Loop spec
+
+- Archetype: spec-satisfaction repair; the verifier is the parity suite plus `rg`.
+- Write scope: `src/providers/registry.ts`, the covering tests, this devlog unit. One
+  branch, one PR. No `main`, no force-push to `dev`.
+- Escalation: if a provider's list turns out to be live-discovery-only such that seeding
+  is meaningless, exclude it and record that here rather than padding the diff.
diff --git a/src/providers/command-code-efforts.ts b/src/providers/command-code-efforts.ts
index b7d5b15d1b..b790c8779d 100644
--- a/src/providers/command-code-efforts.ts
+++ b/src/providers/command-code-efforts.ts
@@ -9,13 +9,6 @@ const COMMAND_CODE_MODEL_EFFORTS = {
     efforts: ["high", "max"],
     profileUrl: "https://commandcode.ai/models/deepseek-v4-flash",
   },
-  // Ox Alpha (stealth preview, added in Command Code v1.31.0): free 1M-context
-  // reasoning model on every plan. The profile does not publish an effort ladder,
-  // so mirror the OpenRouter contract (reasoning mandatory; max/high/low).
-  "stealth/ox-alpha": {
-    efforts: ["low", "high", "max"],
-    profileUrl: "https://commandcode.ai/models/ox-alpha",
-  },
   // Keys must match the EXACT upstream /provider/v1/models ids (GLM ships as
   // `zai-org/GLM-5.3`, not `zai-org/glm-5.3`). The table doubles as the router's
   // known-ids decode source (via `knownModelIdsForProvider`), so a case mismatch
diff --git a/src/providers/registry.ts b/src/providers/registry.ts
index 913ca82af1..40feec0bfd 100644
--- a/src/providers/registry.ts
+++ b/src/providers/registry.ts
@@ -471,13 +471,13 @@ const OPENCODE_GO_THINKING_TOGGLE_MODELS = [
  * images through the proxy's vision sidecar (src/codex/catalog/provider-fetch.ts), a claim nobody
  * has verified for BigModel-hosted GLM.
  */
-const ZHIPU_BIGMODEL_TEXT_MODELS = ["glm-4.6", "glm-4.7", "glm-4.7-flash", "glm-5", "glm-5.1", "glm-5.2", "glm-5.3"];
+const ZHIPU_BIGMODEL_TEXT_MODELS = ["glm-4.6", "glm-4.7", "glm-4.7-flash", "glm-5", "glm-5.1", "glm-5.2", "glm-5.3", "glm-5.3-flash"];
 const ZHIPU_BIGMODEL_MODELS = [...ZHIPU_BIGMODEL_TEXT_MODELS, "glm-4.6v"];
 const ZHIPU_BIGMODEL_INPUT_MODALITIES: Record<string, string[]> = {
   ...Object.fromEntries(ZHIPU_BIGMODEL_TEXT_MODELS.map(id => [id, ["text"]])),
   "glm-4.6v": ["text", "image"],
 };
-const ZHIPU_BIGMODEL_THINKING_TOGGLE_MODELS = ["glm-4.6", "glm-4.7", "glm-5", "glm-5.1", "glm-5.2", "glm-5.3"];
+const ZHIPU_BIGMODEL_THINKING_TOGGLE_MODELS = ["glm-4.6", "glm-4.7", "glm-5", "glm-5.1", "glm-5.2", "glm-5.3", "glm-5.3-flash"];
 const THINKING_BUDGET_EFFORTS = ["low", "medium", "high", "xhigh", "max"];
 // Qwen3.8-Max is the first Qwen3.x model with official direct `reasoning_effort` support.
 // Evidence: https://qwen.ai/blog?id=qwen3.8
@@ -505,8 +505,6 @@ const DEEPSEEK_VISION_PREVIEW_MODEL = "deepseek-v4-flash-vision-exp";
  * capability intersection trusts this map.
  */
 const COMMAND_CODE_IMAGE_MODELS = [
-  "stealth/ox-alpha",
-  "openai/ox-alpha",
   `deepseek/${DEEPSEEK_VISION_PREVIEW_MODEL}`,
   "gpt-5.6-luna",
   "gpt-5.6-sol",
@@ -518,15 +516,6 @@ const COMMAND_CODE_IMAGE_MODELS = [
 const COMMAND_CODE_MODEL_INPUT_MODALITIES: Record<string, ["text", "image"]> =
   Object.fromEntries(COMMAND_CODE_IMAGE_MODELS.map(id => [id, ["text", "image"]]));
 const OPENCODE_FREE_DEEPSEEK_MODELS = ["deepseek-v4-flash-free"];
-/*
- * OpenCode Zen's free slug for the OpenRouter stealth model "Ox Alpha"
- * (openrouter.ai/stealth/ox-alpha): 1,048,576-token context, multimodal
- * (text+image+video upstream; Zen serves text+image), mandatory reasoning,
- * free during the stealth window. Zen displays it as "Ox Alpha Free" under
- * this exact id (opencode.ai/docs/zen, verified 2026-08-21).
- */
-const OPENCODE_OX_ALPHA_FREE_MODEL = "x-preview-f-free";
-const OX_ALPHA_CONTEXT_WINDOW = 1_048_576;
 /*
  * Zen free models that reject `image_url` upstream (#1043, and the reproducible
  * half of #1024).
@@ -613,7 +602,7 @@ const deepseekReasoningMapFor = (modelId: string): Record<string, string> =>
 //           https://help.aliyun.com/en/model-studio/token-plan-quickstart
 const ALIBABA_TOKEN_PLAN_MODELS = [
   "qwen3.8-max", "qwen3.7-max", "qwen3.7-plus", "qwen3.6-flash",
-  "glm-5.3", "glm-5.2", "deepseek-v4-pro",
+  "glm-5.3", "glm-5.3-flash", "glm-5.2", "deepseek-v4-pro",
 ];
 const ALIBABA_TOKEN_PLAN_QWEN_MODELS = [
   "qwen3.8-max", "qwen3.7-max", "qwen3.7-plus", "qwen3.6-flash",
@@ -624,6 +613,7 @@ const ALIBABA_TOKEN_PLAN_INPUT_MODALITIES: Record<string, string[]> = {
   "qwen3.7-plus": ["text", "image"],
   "qwen3.6-flash": ["text", "image"],
   "glm-5.3": ["text"],
+  "glm-5.3-flash": ["text", "image"],
   "glm-5.2": ["text"],
   "deepseek-v4-pro": ["text"],
 };
@@ -636,7 +626,7 @@ const ALIBABA_INTL_TOKEN_PLAN_MODELS = [
   "qwen3.8-max", "qwen3.7-max", "qwen3.7-plus", "qwen3.6-plus", "qwen3.6-flash",
   "deepseek-v4-pro", "deepseek-v4-flash", "deepseek-v3.2",
   "kimi-k2.7-code", "kimi-k2.6", "kimi-k2.5",
-  "glm-5.3", "glm-5.2", "glm-5.1", "glm-5",
+  "glm-5.3", "glm-5.3-flash", "glm-5.2", "glm-5.1", "glm-5",
   "MiniMax-M2.5",
 ];
 const ALIBABA_INTL_TOKEN_PLAN_QWEN_MODELS = [
@@ -686,6 +676,7 @@ const VOLCENGINE_CODING_PLAN_MODELS = [
   "deepseek-v4-pro",
   "deepseek-v4-flash",
   "glm-5.3",
+  "glm-5.3-flash",
   "glm-5.2",
   "kimi-k2.6",
   "minimax-m3",
@@ -694,6 +685,7 @@ const VOLCENGINE_AGENT_PLAN_MODELS = [
   "deepseek-v4-pro",
   "deepseek-v4-flash",
   "glm-5.3",
+  "glm-5.3-flash",
   "glm-5.2",
   "kimi-k2.6",
   "minimax-m3",
@@ -712,6 +704,7 @@ const VOLCENGINE_PLAN_TEXT_ONLY_MODELS = [
   "deepseek-v4-pro",
   "deepseek-v4-flash",
   "glm-5.3",
+  "glm-5.3-flash",
   "glm-5.2",
   "doubao-seed-2.0-pro",
 ];
@@ -728,6 +721,7 @@ const ALIBABA_INTL_TOKEN_PLAN_INPUT_MODALITIES: Record<string, string[]> = {
   "kimi-k2.6": ["text", "image"],
   "kimi-k2.5": ["text", "image"],
   "glm-5.3": ["text"],
+  "glm-5.3-flash": ["text", "image"],
   "glm-5.2": ["text"],
   "glm-5.1": ["text"],
   "glm-5": ["text"],
@@ -851,7 +845,7 @@ const NVIDIA_NIM_NO_VISION_MODELS = [
   "nvidia/nemotron-3-ultra-550b-a55b", "nvidia/nemotron-mini-4b-instruct",
   "nvidia/nvidia-nemotron-nano-9b-v2",
   "openai/gpt-oss-120b", "openai/gpt-oss-20b",
-  "poolside/laguna-xs-2.1", "z-ai/glm-5.3", "z-ai/glm-5.2",
+  "poolside/laguna-xs-2.1", "z-ai/glm-5.3", "z-ai/glm-5.3-flash", "z-ai/glm-5.2",
 ];
 const KIMI_CODING_MODEL_CONTEXT_WINDOWS: Record<string, number> = Object.fromEntries(
   KIMI_CODING_MODELS.map(id => [id, id === "k3[1m]" ? KIMI_K3_1M_CONTEXT_WINDOW : KIMI_K3_STANDARD_CONTEXT_WINDOW]),
@@ -860,7 +854,7 @@ const KIMI_CODING_MODEL_INPUT_MODALITIES = Object.fromEntries(
   KIMI_CODING_K3_MODELS.map(id => [id, ["text", "image"]]),
 );
 const NEURALWATT_REASONING_HISTORY_MODELS = [
-  "glm-5.3", "glm-5.3-short",
+  "glm-5.3", "glm-5.3-short", "glm-5.3-flash",
   "glm-5.2", "glm-5.2-short",
   "kimi-k2.6", "kimi-k2.7-code",
   "qwen3.5-397b", "qwen3.6-35b",
@@ -941,6 +935,7 @@ const DIGITALOCEAN_CHAT_COMPLETION_MODELS = [
   "nemotron-nano-12b-v2-vl",
   "mimo-v2.5-pro",
   "glm-5.3",
+  "glm-5.3-flash",
   "glm-5.2",
   "glm-5.1",
   "glm-5",
@@ -949,6 +944,7 @@ const DIGITALOCEAN_CHAT_COMPLETION_MODELS = [
 ] as const;
 const SCALEWAY_SERVERLESS_CHAT_MODELS = [
   "glm-5.3",
+  "glm-5.3-flash",
   "glm-5.2",
   // gpt-oss-120b is intentionally omitted: Scaleway requires Responses API for tool calling,
   // while this preset routes Codex agent tools through Chat Completions.
@@ -970,6 +966,7 @@ const UMANS_MODELS = [
   "umans-kimi-k2.7",
   "umans-flash",
   "umans-glm-5.3",
+  "umans-glm-5.3-flash",
   "umans-glm-5.2",
   "umans-glm-5.1",
   "umans-qwen3.6-35b-a3b",
@@ -979,12 +976,15 @@ const UMANS_GLM_REASONING_EFFORTS = ["high", "xhigh", "max"];
 // 260814: Z.AI folds GLM-5.3 efforts into low/high/max, so `low` is a real tier here and
 // `xhigh` is not distinct from `max` (docs.z.ai/devpack/latest-model).
 const UMANS_GLM_53_REASONING_EFFORTS = ["low", "high", "max"];
-const UMANS_TEXT_ONLY_MODELS = ["umans-glm-5.3", "umans-glm-5.2", "umans-glm-5.1"];
+const UMANS_TEXT_ONLY_MODELS = ["umans-glm-5.3", "umans-glm-5.3-flash", "umans-glm-5.2", "umans-glm-5.1"];
 const UMANS_MODEL_CONTEXT_WINDOWS: Record<string, number> = {
   "umans-coder": 262_144,
   "umans-kimi-k2.7": 262_144,
   "umans-flash": 262_144,
   "umans-glm-5.3": 405_504,
+  // Mirrors the sibling this provider already carries. Umans has not published a
+  // separate window for the flash tier; asserting a different number would be a guess.
+  "umans-glm-5.3-flash": 405_504,
   "umans-glm-5.2": 405_504,
   "umans-glm-5.1": 202_752,
   "umans-qwen3.6-35b-a3b": 262_144,
@@ -994,6 +994,7 @@ const UMANS_MODEL_INPUT_MODALITIES: Record<string, string[]> = Object.fromEntrie
 );
 const CLINE_PASS_MODELS = [
   "cline-pass/glm-5.3",
+  "cline-pass/glm-5.3-flash",
   "cline-pass/glm-5.2",
   "cline-pass/kimi-k3",
   "cline-pass/kimi-k2.7-code",
@@ -1009,6 +1010,7 @@ const CLINE_PASS_MODELS = [
 ];
 const CLINE_PASS_MODEL_CONTEXT_WINDOWS: Record<string, number> = {
   "cline-pass/glm-5.3": 1_048_576,
+  "cline-pass/glm-5.3-flash": 1_048_576,
   "cline-pass/glm-5.2": 1_048_576,
   "cline-pass/kimi-k3": 1_048_576,
   "cline-pass/kimi-k2.7-code": 262_144,
@@ -1196,11 +1198,9 @@ export const PROVIDER_REGISTRY: readonly ProviderRegistryEntry[] = [
     // Unknown/new live models deliberately do not advertise a reasoning picker.
     reasoningEfforts: [],
     modelReasoningEfforts: COMMAND_CODE_MODEL_REASONING_EFFORTS,
-    // Ox Alpha (stealth preview, changelog v1.31.0): free 1M multimodal reasoning
-    // model on every plan. DeepSeek vision preview id is preemptive metadata —
-    // it is expected to merge into deepseek-v4-flash later.
+    // The DeepSeek vision preview id is preemptive metadata — it is expected to
+    // merge into deepseek-v4-flash later.
     modelContextWindows: {
-      "stealth/ox-alpha": OX_ALPHA_CONTEXT_WINDOW,
       [`deepseek/${DEEPSEEK_VISION_PREVIEW_MODEL}`]: 1_048_576,
     },
     modelInputModalities: COMMAND_CODE_MODEL_INPUT_MODALITIES,
@@ -1373,6 +1373,7 @@ export const PROVIDER_REGISTRY: readonly ProviderRegistryEntry[] = [
       "umans-kimi-k2.7": UMANS_REASONING_EFFORTS,
       "umans-flash": UMANS_REASONING_EFFORTS,
       "umans-glm-5.3": UMANS_GLM_53_REASONING_EFFORTS,
+      "umans-glm-5.3-flash": UMANS_GLM_53_REASONING_EFFORTS,
       "umans-glm-5.2": UMANS_GLM_REASONING_EFFORTS,
       "umans-glm-5.1": UMANS_GLM_REASONING_EFFORTS,
       "umans-qwen3.6-35b-a3b": UMANS_REASONING_EFFORTS,
@@ -1398,21 +1399,19 @@ export const PROVIDER_REGISTRY: readonly ProviderRegistryEntry[] = [
     modelWireDefaults: { "gpt-5.6-luna": "openai-responses", "muse-spark-1.2-contributor": "openai-responses" },
     modelContextWindows: {
       "kimi-k3": KIMI_K3_STANDARD_CONTEXT_WINDOW,
-      // Ox Alpha (stealth 1M multimodal) and the DeepSeek vision preview are
-      // metadata-only here: the Go roster is discovered live, so these apply
-      // the moment the gateway starts serving the ids.
-      [OPENCODE_OX_ALPHA_FREE_MODEL]: OX_ALPHA_CONTEXT_WINDOW,
+      // The DeepSeek vision preview id is metadata-only here: the Go roster is
+      // discovered live, so it applies the moment the gateway serves the id.
       [DEEPSEEK_VISION_PREVIEW_MODEL]: 1_048_576,
     },
     modelInputModalities: {
       "kimi-k3": ["text", "image"],
-      [OPENCODE_OX_ALPHA_FREE_MODEL]: ["text", "image"],
       // Experimental DeepSeek vision preview — expected to merge into deepseek-v4-flash later.
       [DEEPSEEK_VISION_PREVIEW_MODEL]: ["text", "image"],
     },
     modelReasoningEfforts: {
       "gpt-5.6-luna": OPENAI_API_GPT56_REASONING_EFFORTS,
       "glm-5.3": ZAI_GLM_53_REASONING_EFFORTS,
+      "glm-5.3-flash": ZAI_GLM_53_REASONING_EFFORTS,
       "glm-5.2": ZAI_GLM_52_REASONING_EFFORTS,
       "qwen3.8-max": QWEN38_REASONING_EFFORTS,
       "kimi-k3": KIMI_CODING_K3_REASONING_EFFORTS,
@@ -1432,6 +1431,7 @@ export const PROVIDER_REGISTRY: readonly ProviderRegistryEntry[] = [
     },
     modelSupportsReasoningSummaries: {
       "glm-5.3": true,
+      "glm-5.3-flash": true,
       "glm-5.2": true,
       "glm-5.1": true,
       "glm-5": true,
@@ -1455,7 +1455,7 @@ export const PROVIDER_REGISTRY: readonly ProviderRegistryEntry[] = [
     noPenaltyModels: ["kimi-k3", "kimi-k2.7-code", "kimi-k2.7-code-highspeed"],
     autoToolChoiceOnlyModels: ["kimi-k2.7-code", "kimi-k2.7-code-highspeed"],
     // Issue #78: DeepSeek V4 thinking mode requires reasoning_content replay on tool-call turns.
-    preserveReasoningContentModels: ["glm-5.3", "glm-5.2", "kimi-k3", "kimi-k2.7-code", "kimi-k2.7-code-highspeed", ...DEEPSEEK_THINKING_MODELS],
+    preserveReasoningContentModels: ["glm-5.3", "glm-5.3-flash", "glm-5.2", "kimi-k3", "kimi-k2.7-code", "kimi-k2.7-code-highspeed", ...DEEPSEEK_THINKING_MODELS],
   },
   {
     id: "neuralwatt",
@@ -1471,6 +1471,7 @@ export const PROVIDER_REGISTRY: readonly ProviderRegistryEntry[] = [
     // Evidence: devlog/_plan/260710_provider_hardening/003_research_aggregators.md and https://api.neuralwatt.com/v1/models.
     models: [
       "glm-5.3", "glm-5.3-fast", "glm-5.3-short", "glm-5.3-short-fast",
+      "glm-5.3-flash",
       "glm-5.2", "glm-5.2-fast", "glm-5.2-short", "glm-5.2-short-fast",
       "kimi-k2.6", "kimi-k2.6-fast",
       "kimi-k2.7-code",
@@ -1482,6 +1483,9 @@ export const PROVIDER_REGISTRY: readonly ProviderRegistryEntry[] = [
       "glm-5.3-fast": [],
       "glm-5.3-short": ZAI_GLM_53_REASONING_EFFORTS,
       "glm-5.3-short-fast": [],
+      // No `-fast`/`-short` variants are asserted for the flash tier: those suffixes
+      // encode routing Neuralwatt documents per model, and this seed has no source for them.
+      "glm-5.3-flash": ZAI_GLM_53_REASONING_EFFORTS,
       "glm-5.2": ZAI_GLM_52_REASONING_EFFORTS,
       "glm-5.2-fast": [],
       "glm-5.2-short": ZAI_GLM_52_REASONING_EFFORTS,
@@ -1514,16 +1518,11 @@ export const PROVIDER_REGISTRY: readonly ProviderRegistryEntry[] = [
     featured: true,
     dashboardUrl: "https://openrouter.ai/keys",
     jawcodeBundle: "openrouter",
-    // stealth/ox-alpha: free stealth-window frontier model (launched 2026-08-20).
-    // /api/v1/models reports 1,048,576 context, 131,072 max output, text+image+video
-    // input, $0 pricing, mandatory reasoning. Single provider slug: `stealth`.
-    models: ["anthropic/claude-sonnet-5", "stealth/ox-alpha", ...OPENROUTER_GPT56_MODELS],
+    models: ["anthropic/claude-sonnet-5", ...OPENROUTER_GPT56_MODELS],
     modelContextWindows: {
       "anthropic/claude-sonnet-5": 1_000_000,
-      "stealth/ox-alpha": OX_ALPHA_CONTEXT_WINDOW,
       ...OPENROUTER_GPT56_CONTEXT_WINDOWS,
     },
-    modelInputModalities: { "stealth/ox-alpha": ["text", "image"] },
     // OpenRouter documents priority support for OpenAI endpoints, but not Anthropic. Keep the
     // provider unclassified and opt in only the exact OpenAI-backed slugs we ship. These facts
     // belong only to the canonical destination; a same-named custom gateway is unknown to us.
@@ -1934,11 +1933,9 @@ export const PROVIDER_REGISTRY: readonly ProviderRegistryEntry[] = [
     // slash ids — so a Codex-facing slug like `commandcode/deepseek-deepseek-v4-pro`
     // is sent upstream verbatim and rejected with `unsupported_model`.
     modelReasoningEfforts: COMMAND_CODE_MODEL_REASONING_EFFORTS,
-    // Ox Alpha (stealth preview, Command Code changelog v1.31.0) ships with a
-    // 1.05M-token multimodal context; the DeepSeek vision preview id is
-    // preemptive for when the catalog serves it (merges into v4-flash later).
+    // The DeepSeek vision preview id is preemptive for when the catalog serves it
+    // (merges into v4-flash later).
     modelContextWindows: {
-      "stealth/ox-alpha": OX_ALPHA_CONTEXT_WINDOW,
       [`deepseek/${DEEPSEEK_VISION_PREVIEW_MODEL}`]: 1_048_576,
     },
     modelInputModalities: COMMAND_CODE_MODEL_INPUT_MODALITIES,
@@ -2198,8 +2195,8 @@ export const PROVIDER_REGISTRY: readonly ProviderRegistryEntry[] = [
     id: "zai", label: "Z.AI — GLM Coding Plan", baseUrl: "https://api.z.ai/api/coding/paas/v4", adapter: "openai-chat", authKind: "key",
     dashboardUrl: "https://z.ai/manage-apikey/apikey-list", defaultModel: "glm-5.3",
     note: "GLM-5.3 coding subscription",
-    models: ["glm-5.3", "glm-5.3[1m]", "glm-5.2", "glm-5.2[1m]", "glm-5.1", "glm-5", "glm-4.6"],
-    modelContextWindows: { "glm-5.3": 1_000_000, "glm-5.3[1m]": 1_000_000, "glm-5.2": 1_000_000, "glm-5.2[1m]": 1_000_000 },
+    models: ["glm-5.3", "glm-5.3[1m]", "glm-5.3-flash", "glm-5.2", "glm-5.2[1m]", "glm-5.1", "glm-5", "glm-4.6"],
+    modelContextWindows: { "glm-5.3": 1_000_000, "glm-5.3[1m]": 1_000_000, "glm-5.3-flash": 1_000_000, "glm-5.2": 1_000_000, "glm-5.2[1m]": 1_000_000 },
     // Z.AI's OpenAI path returns 400 code 1211 for bracketed model ids.
     modelSuffixBracketStrip: true,
     noVisionModels: ZAI_GLM_5X_MODELS,
@@ -2279,9 +2276,9 @@ export const PROVIDER_REGISTRY: readonly ProviderRegistryEntry[] = [
     authKind: "key",
     dashboardUrl: "https://bigmodel.cn/console/usercenter/apikeys",
     defaultModel: "glm-5.3",
-    models: ["glm-5.3", "glm-5.3[1m]", "glm-5.2", "glm-5.2[1m]", "glm-5.1", "glm-5", "glm-4.6"],
+    models: ["glm-5.3", "glm-5.3[1m]", "glm-5.3-flash", "glm-5.2", "glm-5.2[1m]", "glm-5.1", "glm-5", "glm-4.6"],
     jawcodeBundle: "zai",
-    modelContextWindows: { "glm-5.3": 1_000_000, "glm-5.3[1m]": 1_000_000, "glm-5.2": 1_000_000, "glm-5.2[1m]": 1_000_000 },
+    modelContextWindows: { "glm-5.3": 1_000_000, "glm-5.3[1m]": 1_000_000, "glm-5.3-flash": 1_000_000, "glm-5.2": 1_000_000, "glm-5.2[1m]": 1_000_000 },
     modelSuffixBracketStrip: true,
     noVisionModels: ZAI_GLM_5X_MODELS,
     modelReasoningEfforts: ZAI_GLM_5X_REASONING_EFFORTS,
@@ -2425,12 +2422,13 @@ export const PROVIDER_REGISTRY: readonly ProviderRegistryEntry[] = [
     modelInputModalities: ALIBABA_TOKEN_PLAN_INPUT_MODALITIES,
     modelContextWindows: {
       "qwen3.8-max": 983_616, "qwen3.7-max": 1_000_000, "qwen3.7-plus": 1_000_000,
-      "qwen3.6-flash": 1_000_000, "glm-5.3": 1_000_000, "glm-5.2": 1_000_000, "deepseek-v4-pro": 1_000_000,
+      "qwen3.6-flash": 1_000_000, "glm-5.3": 1_000_000, "glm-5.3-flash": 1_000_000, "glm-5.2": 1_000_000, "deepseek-v4-pro": 1_000_000,
     },
     modelReasoningEfforts: {
       ...Object.fromEntries(ALIBABA_TOKEN_PLAN_QWEN_MODELS.map(id => [id, THINKING_BUDGET_EFFORTS])),
       "qwen3.8-max": QWEN38_REASONING_EFFORTS,
       "glm-5.3": ZAI_GLM_53_REASONING_EFFORTS,
+      "glm-5.3-flash": ZAI_GLM_53_REASONING_EFFORTS,
       "glm-5.2": ZAI_GLM_52_REASONING_EFFORTS,
       "deepseek-v4-pro": deepseekThinkingEffortsFor("deepseek-v4-pro"),
     },
@@ -2438,7 +2436,7 @@ export const PROVIDER_REGISTRY: readonly ProviderRegistryEntry[] = [
     modelReasoningEffortMap: { "deepseek-v4-pro": deepseekReasoningMapFor("deepseek-v4-pro") },
     directReasoningEffortModels: ["qwen3.8-max"],
     thinkingBudgetModels: ALIBABA_TOKEN_PLAN_QWEN_MODELS.filter(id => id !== "qwen3.8-max"),
-    preserveReasoningContentModels: ["glm-5.3", "glm-5.2", "deepseek-v4-pro", "qwen3.8-max", "qwen3.7-max", "qwen3.7-plus", "qwen3.6-flash"],
+    preserveReasoningContentModels: ["glm-5.3", "glm-5.3-flash", "glm-5.2", "deepseek-v4-pro", "qwen3.8-max", "qwen3.7-max", "qwen3.7-plus", "qwen3.6-flash"],
     noVisionModels: ["glm-5.3", "glm-5.2", "deepseek-v4-pro"],
   },
   {
@@ -2461,13 +2459,14 @@ export const PROVIDER_REGISTRY: readonly ProviderRegistryEntry[] = [
       "qwen3.7-max": 1_000_000, "qwen3.7-plus": 1_000_000, "qwen3.6-plus": 1_000_000, "qwen3.6-flash": 1_000_000,
       "deepseek-v4-pro": 1_000_000, "deepseek-v4-flash": 1_000_000, "deepseek-v3.2": 131_072,
       "kimi-k2.7-code": 262_144, "kimi-k2.6": 262_144, "kimi-k2.5": 262_144,
-      "glm-5.3": 1_000_000, "glm-5.2": 1_000_000, "glm-5.1": 1_000_000, "glm-5": 1_000_000,
+      "glm-5.3": 1_000_000, "glm-5.3-flash": 1_000_000, "glm-5.2": 1_000_000, "glm-5.1": 1_000_000, "glm-5": 1_000_000,
       "MiniMax-M2.5": 204_800,
     },
     modelReasoningEfforts: {
       ...Object.fromEntries(ALIBABA_INTL_TOKEN_PLAN_QWEN_MODELS.map(id => [id, THINKING_BUDGET_EFFORTS])),
       "qwen3.8-max": QWEN38_REASONING_EFFORTS,
       "glm-5.3": ZAI_GLM_53_REASONING_EFFORTS,
+      "glm-5.3-flash": ZAI_GLM_53_REASONING_EFFORTS,
       "glm-5.2": ZAI_GLM_52_REASONING_EFFORTS,
       "deepseek-v4-pro": deepseekThinkingEffortsFor("deepseek-v4-pro"),
       "deepseek-v4-flash": deepseekThinkingEffortsFor("deepseek-v4-flash"),
@@ -2478,7 +2477,7 @@ export const PROVIDER_REGISTRY: readonly ProviderRegistryEntry[] = [
     },
     directReasoningEffortModels: ["qwen3.8-max"],
     thinkingBudgetModels: ALIBABA_INTL_TOKEN_PLAN_QWEN_MODELS.filter(id => id !== "qwen3.8-max"),
-    preserveReasoningContentModels: ["glm-5.3", "glm-5.2", "deepseek-v4-pro", "deepseek-v4-flash", "qwen3.8-max", "qwen3.7-max", "qwen3.7-plus", "qwen3.6-plus", "qwen3.6-flash"],
+    preserveReasoningContentModels: ["glm-5.3", "glm-5.3-flash", "glm-5.2", "deepseek-v4-pro", "deepseek-v4-flash", "qwen3.8-max", "qwen3.7-max", "qwen3.7-plus", "qwen3.6-plus", "qwen3.6-flash"],
     noVisionModels: ["deepseek-v4-pro", "deepseek-v4-flash", "deepseek-v3.2", "glm-5.3", "glm-5.2", "glm-5.1", "glm-5", "MiniMax-M2.5"],
     noReasoningModels: ["kimi-k2.7-code", "kimi-k2.6", "kimi-k2.5", "deepseek-v3.2", "glm-5.1", "glm-5", "MiniMax-M2.5"],
     modelDefaultReasoningEfforts: { "qwen3.8-max": "xhigh" },
@@ -2511,10 +2510,10 @@ export const PROVIDER_REGISTRY: readonly ProviderRegistryEntry[] = [
     authKind: "key",
     dashboardUrl: "https://ollama.com/settings/keys",
     // Live IDs verified 2026-07-10; qwen3-coder:480b retires 2026-07-15.
-    models: ["glm-5.3", "glm-5.2", "deepseek-v4-pro", "qwen3-coder:480b", "gpt-oss:120b", "kimi-k2.6", "minimax-m3", "qwen3.5:397b", "gemma4:31b"],
+    models: ["glm-5.3", "glm-5.3-flash", "glm-5.2", "deepseek-v4-pro", "qwen3-coder:480b", "gpt-oss:120b", "kimi-k2.6", "minimax-m3", "qwen3.5:397b", "gemma4:31b"],
     defaultModel: "glm-5.3",
     noVisionModels: [
-      "glm-5.3", "glm-5.2", "glm-5.1", "glm-5", "glm-4.7",
+      "glm-5.3", "glm-5.3-flash", "glm-5.2", "glm-5.1", "glm-5", "glm-4.7",
       "minimax-m2.7", "minimax-m2.5", "minimax-m2.1",
       "nemotron-3-ultra", "nemotron-3-super",
       "deepseek-v4-pro", "deepseek-v4-flash",
@@ -2586,14 +2585,12 @@ export const PROVIDER_REGISTRY: readonly ProviderRegistryEntry[] = [
       [...DEEPSEEK_THINKING_MODELS, ...OPENCODE_FREE_DEEPSEEK_MODELS].map(id => [id, deepseekReasoningMapFor(id)]),
     ),
     preserveReasoningContentModels: [...DEEPSEEK_THINKING_MODELS, ...OPENCODE_FREE_DEEPSEEK_MODELS],
-    // Same Zen gateway as opencode-free: Ox Alpha Free (1M multimodal stealth model)
-    // and the DeepSeek vision preview (merges into deepseek-v4-flash later).
+    // Same Zen gateway as opencode-free: the DeepSeek vision preview id
+    // (merges into deepseek-v4-flash later).
     modelContextWindows: {
-      [OPENCODE_OX_ALPHA_FREE_MODEL]: OX_ALPHA_CONTEXT_WINDOW,
       [DEEPSEEK_VISION_PREVIEW_MODEL]: 1_048_576,
     },
     modelInputModalities: {
-      [OPENCODE_OX_ALPHA_FREE_MODEL]: ["text", "image"],
       [DEEPSEEK_VISION_PREVIEW_MODEL]: ["text", "image"],
     },
     noVisionModels: [...OPENCODE_ZEN_TEXT_ONLY_MODELS, ...DEEPSEEK_THINKING_MODELS],
@@ -2626,16 +2623,12 @@ export const PROVIDER_REGISTRY: readonly ProviderRegistryEntry[] = [
     modelReasoningEfforts: Object.fromEntries(OPENCODE_FREE_DEEPSEEK_MODELS.map(id => [id, deepseekThinkingEffortsFor(id)])),
     modelReasoningEffortMap: Object.fromEntries(OPENCODE_FREE_DEEPSEEK_MODELS.map(id => [id, deepseekReasoningMapFor(id)])),
     preserveReasoningContentModels: OPENCODE_FREE_DEEPSEEK_MODELS,
-    // Ox Alpha Free (`x-preview-f-free`): the OpenRouter stealth model on Zen's
-    // free tier — 1,048,576 context, text+image input. Deliberately NOT in the
-    // text-only list below. The DeepSeek vision preview id is preemptive
-    // metadata for when Zen starts serving it (merges into v4-flash later).
+    // The DeepSeek vision preview id is preemptive metadata for when Zen starts
+    // serving it (merges into v4-flash later).
     modelContextWindows: {
-      [OPENCODE_OX_ALPHA_FREE_MODEL]: OX_ALPHA_CONTEXT_WINDOW,
       [DEEPSEEK_VISION_PREVIEW_MODEL]: 1_048_576,
     },
     modelInputModalities: {
-      [OPENCODE_OX_ALPHA_FREE_MODEL]: ["text", "image"],
       [DEEPSEEK_VISION_PREVIEW_MODEL]: ["text", "image"],
     },
     // Same Zen roster behind the same base URL, so it carries the same measured
@@ -2727,6 +2720,7 @@ export const PROVIDER_REGISTRY: readonly ProviderRegistryEntry[] = [
       "@cf/deepseek-ai/deepseek-r1-distill-qwen-32b",
       "@cf/moonshotai/kimi-k2.7-code",
       "@cf/zai-org/glm-5.3",
+      "@cf/zai-org/glm-5.3-flash",
       "@cf/zai-org/glm-5.2",
       "@cf/mistralai/mistral-small-3.1-24b-instruct",
     ],
diff --git a/tests/alibaba-intl-token-plan.test.ts b/tests/alibaba-intl-token-plan.test.ts
index 459ccfd092..c76a9c364b 100644
--- a/tests/alibaba-intl-token-plan.test.ts
+++ b/tests/alibaba-intl-token-plan.test.ts
@@ -32,9 +32,10 @@ describe("alibaba-token-plan-intl registry entry", () => {
     expect(entry!.models).toContain("kimi-k2.7-code");
     expect(entry!.models).toContain("glm-5.2");
     expect(entry!.models).toContain("glm-5.3");
+    expect(entry!.models).toContain("glm-5.3-flash");
     expect(entry!.models).toContain("MiniMax-M2.5");
     expect(entry!.models).toContain("qwen3.8-max");
-    expect(entry!.models!.length).toBe(16);
+    expect(entry!.models!.length).toBe(17);
   });
 
   test("MiniMax case-insensitive normalization is set", () => {
diff --git a/tests/cline-pass-provider.test.ts b/tests/cline-pass-provider.test.ts
index 3bd7d7b827..94072bc60a 100644
--- a/tests/cline-pass-provider.test.ts
+++ b/tests/cline-pass-provider.test.ts
@@ -9,6 +9,7 @@ import type { OcxConfig, OcxParsedRequest } from "../src/types";
 
 const OFFICIAL_CLINE_PASS_MODELS = [
   "cline-pass/glm-5.3",
+  "cline-pass/glm-5.3-flash",
   "cline-pass/glm-5.2",
   "cline-pass/kimi-k3",
   "cline-pass/kimi-k2.7-code",
@@ -71,6 +72,7 @@ describe("ClinePass provider", () => {
     expect(entry?.modelMaxInputTokens).toBeUndefined();
     expect(entry?.noVisionModels).toEqual([
       "cline-pass/glm-5.3",
+  "cline-pass/glm-5.3-flash",
       "cline-pass/glm-5.2",
       "cline-pass/deepseek-v4-pro",
       "cline-pass/deepseek-v4-flash",
diff --git a/tests/command-code-provider.test.ts b/tests/command-code-provider.test.ts
index b342b56da5..9c5e02ed26 100644
--- a/tests/command-code-provider.test.ts
+++ b/tests/command-code-provider.test.ts
@@ -87,8 +87,6 @@ describe("Command Code provider", () => {
     const oauth = PROVIDER_REGISTRY.find(row => row.id === "command-code");
     const apiKey = PROVIDER_REGISTRY.find(row => row.id === "commandcode");
     const verifiedImageModels = [
-      "stealth/ox-alpha",
-      "openai/ox-alpha",
       "deepseek/deepseek-v4-flash-vision-exp",
       "gpt-5.6-luna",
       "gpt-5.6-sol",
diff --git a/tests/provider-model-discovery-contract.test.ts b/tests/provider-model-discovery-contract.test.ts
index c49d4062dd..b8b47adf9e 100644
--- a/tests/provider-model-discovery-contract.test.ts
+++ b/tests/provider-model-discovery-contract.test.ts
@@ -442,6 +442,7 @@ describe("registry-owned provider model discovery", () => {
       "@cf/deepseek-ai/deepseek-r1-distill-qwen-32b",
       "@cf/moonshotai/kimi-k2.7-code",
       "@cf/zai-org/glm-5.3",
+      "@cf/zai-org/glm-5.3-flash",
       "@cf/zai-org/glm-5.2",
       "@cf/mistralai/mistral-small-3.1-24b-instruct",
     ]);
diff --git a/tests/provider-registry-parity.test.ts b/tests/provider-registry-parity.test.ts
index e421907058..1f155d76df 100644
--- a/tests/provider-registry-parity.test.ts
+++ b/tests/provider-registry-parity.test.ts
@@ -307,7 +307,7 @@ describe("provider registry parity", () => {
       liveModels: false,
       models: [
         "qwen3.8-max", "qwen3.7-max", "qwen3.7-plus", "qwen3.6-flash",
-        "glm-5.3", "glm-5.2", "deepseek-v4-pro",
+        "glm-5.3", "glm-5.3-flash", "glm-5.2", "deepseek-v4-pro",
       ],
       modelInputModalities: {
         "qwen3.8-max": ["text", "image"],
@@ -340,6 +340,7 @@ describe("provider registry parity", () => {
     const neuralwatt = PROVIDER_REGISTRY.find(entry => entry.id === "neuralwatt");
     expect(neuralwatt?.models).toEqual([
       "glm-5.3", "glm-5.3-fast", "glm-5.3-short", "glm-5.3-short-fast",
+      "glm-5.3-flash",
       "glm-5.2", "glm-5.2-fast", "glm-5.2-short", "glm-5.2-short-fast",
       "kimi-k2.6", "kimi-k2.6-fast", "kimi-k2.7-code",
       "qwen3.5-397b", "qwen3.5-397b-fast", "qwen3.6-35b", "qwen3.6-35b-fast",
@@ -367,7 +368,7 @@ describe("provider registry parity", () => {
     const optedInProviders = PROVIDER_REGISTRY
       .filter(entry => entry.modelSuffixBracketStrip)
       .map(entry => entry.id);
-    expect(zai?.modelContextWindows).toEqual({ "glm-5.3": 1_000_000, "glm-5.3[1m]": 1_000_000, "glm-5.2": 1_000_000, "glm-5.2[1m]": 1_000_000 });
+    expect(zai?.modelContextWindows).toEqual({ "glm-5.3": 1_000_000, "glm-5.3[1m]": 1_000_000, "glm-5.3-flash": 1_000_000, "glm-5.2": 1_000_000, "glm-5.2[1m]": 1_000_000 });
     expect(zai?.modelDefaultReasoningEfforts).toEqual({ "glm-5.3": "max", "glm-5.3[1m]": "max" });
     expect(zai?.modelMaxOutputTokens).toEqual({ "glm-5.3": 131_072, "glm-5.3[1m]": 131_072 });
     expect(providerConfigSeed(zai!).modelSuffixBracketStrip).toBe(true);
@@ -582,7 +583,7 @@ describe("provider registry parity", () => {
     const ollamaCloud = PROVIDER_REGISTRY.find(entry => entry.id === "ollama-cloud");
 
     expect(ollamaCloud?.models).toEqual([
-      "glm-5.3", "glm-5.2", "deepseek-v4-pro", "qwen3-coder:480b", "gpt-oss:120b",
+      "glm-5.3", "glm-5.3-flash", "glm-5.2", "deepseek-v4-pro", "qwen3-coder:480b", "gpt-oss:120b",
       "kimi-k2.6", "minimax-m3", "qwen3.5:397b", "gemma4:31b",
     ]);
     expect(ollamaCloud?.models).not.toContain("qwen3-coder");
diff --git a/tests/volcengine-providers.test.ts b/tests/volcengine-providers.test.ts
index bc1ee444b9..2dba569a24 100644
--- a/tests/volcengine-providers.test.ts
+++ b/tests/volcengine-providers.test.ts
@@ -60,6 +60,7 @@ describe("Volcengine Ark providers", () => {
         "deepseek-v4-pro",
         "deepseek-v4-flash",
         "glm-5.3",
+        "glm-5.3-flash",
         "glm-5.2",
         "kimi-k2.6",
         "minimax-m3",
@@ -93,6 +94,7 @@ describe("Volcengine Ark providers", () => {
         "deepseek-v4-pro",
         "deepseek-v4-flash",
         "glm-5.3",
+        "glm-5.3-flash",
         "glm-5.2",
         "kimi-k2.6",
         "minimax-m3",