diff --git a/src/codex/catalog/provider-fetch.ts b/src/codex/catalog/provider-fetch.ts
index 565f6057a43..e0e70009fe3 100644
--- a/src/codex/catalog/provider-fetch.ts
+++ b/src/codex/catalog/provider-fetch.ts
@@ -41,7 +41,7 @@ import type { FastPolicyAuthority } from "../../providers/fastwire";
 import { effectiveGoogleMode, getProviderRegistryEntry, providerMatchesRegistryTransport } from "../../providers/registry";
 import { parseAntigravityAvailableModels, registerAntigravityDiscoveredWireModels } from "../../providers/antigravity-models";
 import { applyProviderContextCap, providerContextCap, resolveUnknownRoutedContextWindow } from "../../providers/context-cap";
-import { routedSlug, slugEquals, slugsEquivalent } from "../../providers/slug-codec";
+import { routedSlug, slugEquals, slugEquivalenceKey, slugsEquivalent } from "../../providers/slug-codec";
 import { CODEX_GPT5_IDENTITY_LINE } from "../../adapters/identity";
 import { filterCursorConfiguredModelsByLiveDiscovery } from "../../adapters/cursor/discovery";
 import { fetchCursorUsableModels } from "../../adapters/cursor/live-models";
@@ -1560,7 +1560,24 @@ export function filterCatalogVisibleModels(
   const allowByProvider = new Map<string, Set<string>>();
   for (const [name, prov] of Object.entries(config.providers)) {
     const sel = prov.selectedModels;
-    if (Array.isArray(sel) && sel.length > 0) allowByProvider.set(name, new Set(sel));
+    // Keyed the way `sync.ts` keys the same list, so a slash-bearing native id and
+    // the encoded slug the Codex picker displays are one entry rather than two. A
+    // bare `Set(sel)` matched only the native form, so an allowlist written from the
+    // displayed slug — which `ocx models remove` also accepts — hid every model it
+    // was meant to keep.
+    //
+    // The key is deliberately lossy: `p/a/b` and `p/a-b` collapse to one entry, so a
+    // provider publishing both spellings has them selected together. That is a real
+    // limitation, pinned by the tests below and tracked as a follow-up; it is NOT
+    // fixed here. Resolving selections against the current roster instead was tried
+    // and rejected — the roster is an incomplete dictionary (live discovery can omit
+    // a published id), so it produces the same over-grant while additionally
+    // disagreeing with the `slugEquivalenceKey` contract `sync.ts` uses at merge time.
+    // Two catalog stages with different equivalence relations is the exact bug class
+    // this change exists to remove.
+    if (Array.isArray(sel) && sel.length > 0) {
+      allowByProvider.set(name, new Set(sel.map(model => slugEquivalenceKey(routedSlug(name, model)))));
+    }
   }
   return models.filter(m => {
     const nativeAlias = m.provider === COMBO_NAMESPACE && m.nativeAlias === true;
@@ -1572,7 +1589,7 @@ export function filterCatalogVisibleModels(
       if (slugEquals(stored, m.provider, m.id)) return false;
     }
     const allow = allowByProvider.get(m.provider);
-    return !allow || allow.has(m.id);
+    return !allow || allow.has(slugEquivalenceKey(routedSlug(m.provider, m.id)));
   });
 }
 
diff --git a/tests/selected-models.test.ts b/tests/selected-models.test.ts
index 1e28df11026..755060410cc 100644
--- a/tests/selected-models.test.ts
+++ b/tests/selected-models.test.ts
@@ -47,3 +47,77 @@ describe("filterCatalogVisibleModels — per-provider allowlist", () => {
     expect(out.map(x => x.id).sort()).toEqual(["model-1999", "model-7"]);
   });
 });
+
+describe("filterCatalogVisibleModels — slash-bearing ids", () => {
+  // The Codex picker displays a slash-bearing native id in its ENCODED form, and
+  // `ocx models remove` accepts that form too, so an allowlist is routinely written
+  // with slugs the provider never published. A bare `Set(selectedModels)` matched
+  // only the native spelling and hid every model it was meant to keep.
+  const native = "moonshotai/kimi-k3-free";
+  const encoded = "moonshotai-kimi-k3-free";
+  const rows = [m("zenmux", native), m("zenmux", "openai/gpt-5.5")];
+
+  test("an allowlist written with the encoded slug keeps the model", () => {
+    const visible = filterCatalogVisibleModels(rows, cfg({
+      zenmux: { selectedModels: [encoded] },
+    }));
+    expect(visible.map(v => v.id)).toEqual([native]);
+  });
+
+  test("the native form keeps working", () => {
+    const visible = filterCatalogVisibleModels(rows, cfg({
+      zenmux: { selectedModels: [native] },
+    }));
+    expect(visible.map(v => v.id)).toEqual([native]);
+  });
+
+  test("a mixed allowlist keeps both, without duplicating either", () => {
+    const visible = filterCatalogVisibleModels(rows, cfg({
+      zenmux: { selectedModels: [encoded, "openai/gpt-5.5"] },
+    }));
+    expect(visible.map(v => v.id).sort()).toEqual([native, "openai/gpt-5.5"].sort());
+  });
+
+  test("a model outside the allowlist is still hidden", () => {
+    const visible = filterCatalogVisibleModels(rows, cfg({
+      zenmux: { selectedModels: [encoded] },
+    }));
+    expect(visible.map(v => v.id)).not.toContain("openai/gpt-5.5");
+  });
+
+  // The encoding is lossy: `a/b` and `a-b` share one encoded form, so a provider that
+  // publishes both spellings has them selected TOGETHER. These tests pin that as known
+  // behavior rather than leaving it undiscovered.
+  //
+  // It is not fixed here, and the obvious alternative does not work. Resolving each
+  // selection against the provider's current rows was tried: the roster is an
+  // incomplete dictionary — live discovery can omit a published id — so an exact
+  // `a-b` selection still resolves onto `a/b` whenever `a-b` is missing from that
+  // snapshot, producing the same over-grant. It would additionally disagree with the
+  // `slugEquivalenceKey` relation `sync.ts` applies when merging the persisted
+  // catalog, and two catalog stages using different equivalence rules is the bug
+  // class this change removes. A real fix needs ONE selection resolver shared by
+  // filtering, persisted sync, and routing, against a complete known-id set.
+  describe("a provider publishing both spellings (known lossy case)", () => {
+    const both = [m("p", "a/b"), m("p", "a-b")];
+
+    test("selecting either spelling selects both", () => {
+      const viaSlash = filterCatalogVisibleModels(both, cfg({ p: { selectedModels: ["a/b"] } }));
+      expect(viaSlash.map(v => v.id).sort()).toEqual(["a-b", "a/b"]);
+      const viaDash = filterCatalogVisibleModels(both, cfg({ p: { selectedModels: ["a-b"] } }));
+      expect(viaDash.map(v => v.id).sort()).toEqual(["a-b", "a/b"]);
+    });
+
+    test("an unrelated model is still hidden, so the allowlist has not collapsed", () => {
+      const rows = [...both, m("p", "unrelated")];
+      const visible = filterCatalogVisibleModels(rows, cfg({ p: { selectedModels: ["a-b"] } }));
+      expect(visible.map(v => v.id)).not.toContain("unrelated");
+    });
+  });
+
+  test("a nested-slash id is kept by its fully encoded slug", () => {
+    const nested = [m("p", "x/y/z"), m("p", "other")];
+    const visible = filterCatalogVisibleModels(nested, cfg({ p: { selectedModels: ["x-y-z"] } }));
+    expect(visible.map(v => v.id)).toEqual(["x/y/z"]);
+  });
+});