diff --git a/packages/app/src/components/GlobalFilterContext.tsx b/packages/app/src/components/GlobalFilterContext.tsx
index 2780a202..b7e12e08 100644
--- a/packages/app/src/components/GlobalFilterContext.tsx
+++ b/packages/app/src/components/GlobalFilterContext.tsx
@@ -25,6 +25,7 @@ import {
   Sequence,
   SEQUENCE_OPTIONS,
 } from '@/lib/data-mappings';
+import { computeAutoSwitchDecision } from '@/lib/unofficial-run-auto-switch';
 import type { AvailabilityRow, WorkflowInfoResponse } from '@/lib/api';
 
 interface RunInfo {
@@ -172,6 +173,34 @@ export function GlobalFilterProvider({ children }: { children: ReactNode }) {
     });
   }, [availabilityRows, unofficialAvailable]);
 
+  // Auto-switch the selected model when an unofficial run is loaded that
+  // doesn't include the currently selected model. Without this, navigating
+  // to `?unofficialrun=<id>` while the default `g_model=DeepSeek-R1` sticks
+  // leaves the user staring at a chart with no overlay points — they'd have
+  // to know to open the dropdown and pick the run's model themselves.
+  //
+  // Precedence on first load: the `if (urlModel)` early-bail in
+  // `computeAutoSwitchDecision` is the primary guard for explicit `g_model`
+  // intent. The dedupe ref is a secondary guard for the narrow window after
+  // an auto-switch fires but before the URL-sync effect (below) writes
+  // `g_model` back to the URL — once that runs, `urlModel` is set on every
+  // subsequent render and the ref check is effectively redundant. The ref
+  // still matters across navigations between unofficial runs because it is
+  // reset whenever the overlay set goes empty.
+  const lastAutoSwitchKeyRef = useRef<string>('');
+  useEffect(() => {
+    const decision = computeAutoSwitchDecision(
+      unofficialAvailable,
+      getUrlParam('g_model'),
+      selectedModel,
+      lastAutoSwitchKeyRef.current,
+    );
+    lastAutoSwitchKeyRef.current = decision.nextKey;
+    if (decision.modelToSet !== null) {
+      setSelectedModel(decision.modelToSet);
+    }
+  }, [unofficialAvailable, selectedModel]);
+
   // Sequences available for the selected model (DB ∪ unofficial run for this model)
   const availableSequences = useMemo(() => {
     const unofficialSeqs = unofficialAvailable
diff --git a/packages/app/src/components/unofficial-run-provider.tsx b/packages/app/src/components/unofficial-run-provider.tsx
index 9de84519..fa02bcaf 100644
--- a/packages/app/src/components/unofficial-run-provider.tsx
+++ b/packages/app/src/components/unofficial-run-provider.tsx
@@ -43,7 +43,7 @@ type UnofficialChartData = Record<
 
 const UNOFFICIAL_RUN_PARAM_RE = /^unofficialruns?$/i;
 
-interface AvailableModelSequence {
+export interface AvailableModelSequence {
   model: Model;
   sequence: Sequence;
   precisions: string[];
diff --git a/packages/app/src/lib/unofficial-run-auto-switch.test.ts b/packages/app/src/lib/unofficial-run-auto-switch.test.ts
new file mode 100644
index 00000000..f58776ad
--- /dev/null
+++ b/packages/app/src/lib/unofficial-run-auto-switch.test.ts
@@ -0,0 +1,114 @@
+import { describe, expect, it } from 'vitest';
+
+import type { AvailableModelSequence } from '@/components/unofficial-run-provider';
+import { Model, Sequence } from '@/lib/data-mappings';
+
+import { computeAutoSwitchDecision } from './unofficial-run-auto-switch';
+
+function entry(model: Model, sequence: Sequence): AvailableModelSequence {
+  return { model, sequence, precisions: [] };
+}
+
+describe('computeAutoSwitchDecision', () => {
+  it('returns no-op and resets the key when no unofficial run is loaded', () => {
+    expect(computeAutoSwitchDecision([], undefined, Model.DeepSeek_R1, 'stale-key')).toEqual({
+      nextKey: '',
+      modelToSet: null,
+    });
+  });
+
+  it('switches to the run model when g_model is not pinned and current model is not covered', () => {
+    const run = [entry(Model.DeepSeek_V4_Pro, Sequence.OneK_OneK)];
+    const decision = computeAutoSwitchDecision(run, undefined, Model.DeepSeek_R1, '');
+    expect(decision.modelToSet).toBe(Model.DeepSeek_V4_Pro);
+    expect(decision.nextKey).toBe(Model.DeepSeek_V4_Pro);
+  });
+
+  it('respects an explicit g_model URL param even when the run lacks that model', () => {
+    const run = [entry(Model.DeepSeek_V4_Pro, Sequence.OneK_OneK)];
+    const decision = computeAutoSwitchDecision(run, Model.DeepSeek_R1, Model.DeepSeek_R1, '');
+    expect(decision.modelToSet).toBeNull();
+    // Ref must not be advanced — if the URL is later cleared we still want
+    // a fresh load of the same run to be able to fire the switch.
+    expect(decision.nextKey).toBe('');
+  });
+
+  it('does not switch when the current model is already covered by the overlay', () => {
+    const run = [
+      entry(Model.DeepSeek_R1, Sequence.OneK_OneK),
+      entry(Model.DeepSeek_V4_Pro, Sequence.OneK_OneK),
+    ];
+    const decision = computeAutoSwitchDecision(run, undefined, Model.DeepSeek_R1, '');
+    expect(decision.modelToSet).toBeNull();
+    // Key still advances so we don't keep re-evaluating on every render.
+    expect(decision.nextKey).toBe([Model.DeepSeek_R1, Model.DeepSeek_V4_Pro].toSorted().join(','));
+  });
+
+  it('does not re-fire after a manual model change against the same run set', () => {
+    // Simulate the post-auto-switch state: ref already holds the run's key,
+    // user manually switched back to a model the run does not cover.
+    const run = [entry(Model.DeepSeek_V4_Pro, Sequence.OneK_OneK)];
+    const lastKey = Model.DeepSeek_V4_Pro;
+    const decision = computeAutoSwitchDecision(run, undefined, Model.DeepSeek_R1, lastKey);
+    expect(decision.modelToSet).toBeNull();
+    expect(decision.nextKey).toBe(lastKey);
+  });
+
+  it('re-arms after the overlay set is cleared so a subsequent load can switch again', () => {
+    // Step 1: a run is loaded, switch fires.
+    const run = [entry(Model.DeepSeek_V4_Pro, Sequence.OneK_OneK)];
+    const first = computeAutoSwitchDecision(run, undefined, Model.DeepSeek_R1, '');
+    expect(first.modelToSet).toBe(Model.DeepSeek_V4_Pro);
+
+    // Step 2: user dismisses the run, overlay set goes empty — ref resets.
+    const cleared = computeAutoSwitchDecision([], undefined, Model.DeepSeek_V4_Pro, first.nextKey);
+    expect(cleared).toEqual({ nextKey: '', modelToSet: null });
+
+    // Step 3: a *different* run is loaded with a different model. The cleared
+    // ref allows the switch to fire again.
+    const run2 = [entry(Model.Kimi_K2_5, Sequence.OneK_OneK)];
+    const second = computeAutoSwitchDecision(
+      run2,
+      undefined,
+      Model.DeepSeek_V4_Pro,
+      cleared.nextKey,
+    );
+    expect(second.modelToSet).toBe(Model.Kimi_K2_5);
+  });
+
+  it('ignores sequence-only changes in the dedupe key', () => {
+    // Same model, two sequences appearing across renders. The decision logic
+    // only branches on model, so the key should not change when a new
+    // sequence arrives for an already-covered model — otherwise the effect
+    // would re-evaluate (and bail) on every sequence delta.
+    const oneK = [entry(Model.DeepSeek_V4_Pro, Sequence.OneK_OneK)];
+    const both = [
+      entry(Model.DeepSeek_V4_Pro, Sequence.OneK_OneK),
+      entry(Model.DeepSeek_V4_Pro, Sequence.EightK_OneK),
+    ];
+    const first = computeAutoSwitchDecision(oneK, undefined, Model.DeepSeek_R1, '');
+    const second = computeAutoSwitchDecision(both, undefined, Model.DeepSeek_V4_Pro, first.nextKey);
+    expect(first.nextKey).toBe(second.nextKey);
+    expect(second.modelToSet).toBeNull();
+  });
+
+  it('picks the first model deterministically across insertion orders', () => {
+    // Same set of models in two different orders should produce the same
+    // auto-picked target — protecting against `Object.keys`-driven nondeterminism
+    // in `parseAvailableModelsAndSequences`.
+    const orderA = [
+      entry(Model.MiniMax_M2_5, Sequence.OneK_OneK),
+      entry(Model.DeepSeek_V4_Pro, Sequence.OneK_OneK),
+      entry(Model.Kimi_K2_5, Sequence.OneK_OneK),
+    ];
+    const orderB = [
+      entry(Model.Kimi_K2_5, Sequence.OneK_OneK),
+      entry(Model.DeepSeek_V4_Pro, Sequence.OneK_OneK),
+      entry(Model.MiniMax_M2_5, Sequence.OneK_OneK),
+    ];
+    const a = computeAutoSwitchDecision(orderA, undefined, Model.DeepSeek_R1, '');
+    const b = computeAutoSwitchDecision(orderB, undefined, Model.DeepSeek_R1, '');
+    expect(a.modelToSet).toBe(b.modelToSet);
+    expect(a.nextKey).toBe(b.nextKey);
+  });
+});
diff --git a/packages/app/src/lib/unofficial-run-auto-switch.ts b/packages/app/src/lib/unofficial-run-auto-switch.ts
new file mode 100644
index 00000000..a4af4683
--- /dev/null
+++ b/packages/app/src/lib/unofficial-run-auto-switch.ts
@@ -0,0 +1,48 @@
+import type { AvailableModelSequence } from '@/components/unofficial-run-provider';
+import type { Model } from '@/lib/data-mappings';
+
+export interface AutoSwitchDecision {
+  /** New value the caller should write into the dedupe ref. */
+  nextKey: string;
+  /** Model to switch to, or null when no switch is needed. */
+  modelToSet: Model | null;
+}
+
+/**
+ * Pure decision helper for the unofficial-run auto-switch effect in
+ * `GlobalFilterContext`. Given the unofficial run's available models, the URL
+ * `g_model` param, the currently selected model, and the previous dedupe key,
+ * returns whether to swap `selectedModel` and what the new dedupe key should be.
+ *
+ * - When the overlay set is empty, the dedupe key is reset so the next load
+ *   re-arms the effect.
+ * - When the URL pinned `g_model` explicitly, no switch fires (respect intent).
+ * - Otherwise the dedupe key is the sorted unique list of overlay models — the
+ *   sequence dimension is intentionally excluded so a sequence-only delta does
+ *   not invalidate a manual model pick the user made earlier.
+ * - The first model is taken from a sorted unique list to keep the choice
+ *   deterministic across renders (insertion order from `Object.keys` is not
+ *   guaranteed for multi-model runs).
+ */
+export function computeAutoSwitchDecision(
+  unofficialAvailable: AvailableModelSequence[],
+  urlModel: string | undefined,
+  selectedModel: Model,
+  lastKey: string,
+): AutoSwitchDecision {
+  if (unofficialAvailable.length === 0) {
+    return { nextKey: '', modelToSet: null };
+  }
+  if (urlModel) {
+    return { nextKey: lastKey, modelToSet: null };
+  }
+  const sortedModels = [...new Set(unofficialAvailable.map((a) => a.model))].toSorted();
+  const key = sortedModels.join(',');
+  if (lastKey === key) {
+    return { nextKey: lastKey, modelToSet: null };
+  }
+  if (sortedModels.includes(selectedModel)) {
+    return { nextKey: key, modelToSet: null };
+  }
+  return { nextKey: key, modelToSet: sortedModels[0] };
+}