diff --git a/src/hooks/runtime-fallback/constants.ts b/src/hooks/runtime-fallback/constants.ts
index 19a7cad563..600940b094 100644
--- a/src/hooks/runtime-fallback/constants.ts
+++ b/src/hooks/runtime-fallback/constants.ts
@@ -35,6 +35,9 @@ export const RETRYABLE_ERROR_PATTERNS = [
   /service.?unavailable/i,
   /overloaded/i,
   /temporarily.?unavailable/i,
+  /hit.{0,10}(?:your|the)?.{0,5}limit/i,
+  /spending.?cap/i,
+  /usage.?limit/i,
   /try.?again/i,
   /(?:^|\s)429(?:\s|$)/,
   /(?:^|\s)503(?:\s|$)/,
diff --git a/src/shared/model-error-classifier.test.ts b/src/shared/model-error-classifier.test.ts
index a1f7c52654..92bf01d148 100644
--- a/src/shared/model-error-classifier.test.ts
+++ b/src/shared/model-error-classifier.test.ts
@@ -3,7 +3,7 @@ const { describe, expect, test, beforeEach, afterEach, mock, spyOn } = require("
 import * as connectedProvidersCache from "./connected-providers-cache"
 
 let readConnectedProvidersCacheSpy: ReturnType<typeof spyOn> | undefined
-const { shouldRetryError, selectFallbackProvider } = await import("./model-error-classifier")
+const { shouldRetryError, isRetryableModelError, isStopModelError, selectFallbackProvider } = await import("./model-error-classifier")
 
 describe("model-error-classifier", () => {
   beforeEach(() => {
@@ -16,226 +16,410 @@ describe("model-error-classifier", () => {
     readConnectedProvidersCacheSpy = undefined
   })
 
-  test("treats overloaded retry messages as retryable", () => {
-    //#given
-    const error = { message: "Provider is overloaded" }
+  describe("#shouldRetryError", () => {
+    test("treats overloaded retry messages as retryable", () => {
+      //#given
+      const error = { message: "Provider is overloaded" }
 
-    //#when
-    const result = shouldRetryError(error)
+      //#when
+      const result = shouldRetryError(error)
 
-    //#then
-    expect(result).toBe(true)
-  })
+      //#then
+      expect(result).toBe(true)
+    })
 
-  test("treats cooling-down auto-retry messages as retryable", () => {
-    //#given
-    const error = {
-      message:
-        "All credentials for model claude-opus-4-6-thinking are cooling down [retrying in ~5 days attempt #1]",
-    }
+    test("treats cooling-down auto-retry messages as retryable", () => {
+      //#given
+      const error = {
+        message:
+          "All credentials for model claude-opus-4-6-thinking are cooling down [retrying in ~5 days attempt #1]",
+      }
 
-    //#when
-    const result = shouldRetryError(error)
+      //#when
+      const result = shouldRetryError(error)
 
-    //#then
-    expect(result).toBe(true)
-  })
+      //#then
+      expect(result).toBe(true)
+    })
 
-  test("selectFallbackProvider prefers first connected provider in preference order", () => {
-    //#given
-    readConnectedProvidersCacheSpy?.mockReturnValue(["anthropic", "nvidia"])
+    test("treats 'bad request' message as retryable (GitHub Copilot rolling update)", () => {
+      //#given
+      const error = { message: "400 Bad Request" }
 
-    //#when
-    const provider = selectFallbackProvider(["anthropic", "nvidia"], "nvidia")
+      //#when
+      const result = shouldRetryError(error)
 
-    //#then
-    expect(provider).toBe("anthropic")
-  })
+      //#then
+      expect(result).toBe(true)
+    })
 
-  test("selectFallbackProvider falls back to next connected provider when first is disconnected", () => {
-    //#given
-    readConnectedProvidersCacheSpy?.mockReturnValue(["nvidia"])
+    test("treats 'bad request' lowercase as retryable", () => {
+      //#given
+      const error = { message: "bad request: model temporarily unavailable" }
 
-    //#when
-    const provider = selectFallbackProvider(["anthropic", "nvidia"])
+      //#when
+      const result = shouldRetryError(error)
 
-    //#then
-    expect(provider).toBe("nvidia")
-  })
+      //#then
+      expect(result).toBe(true)
+    })
 
-  test("selectFallbackProvider uses provider preference order when cache is missing", () => {
-    //#given - no cache file
+    test("treats HTTP 429 rate limit message as retryable", () => {
+      //#given
+      const error = { message: "429 Too Many Requests: rate limit reached" }
 
-    //#when
-    const provider = selectFallbackProvider(["anthropic", "nvidia"], "nvidia")
+      //#when
+      const result = shouldRetryError(error)
 
-    //#then
-    expect(provider).toBe("anthropic")
-  })
+      //#then
+      expect(result).toBe(true)
+    })
 
-  test("selectFallbackProvider uses connected preferred provider when fallback providers are unavailable", () => {
-    //#given
-    readConnectedProvidersCacheSpy?.mockReturnValue(["provider-x"])
+    test("triggers fallback for QuotaExceededError (provider exhausted, different provider may work)", () => {
+      //#given
+      const error = { name: "QuotaExceededError" }
 
-    //#when
-    const provider = selectFallbackProvider(["provider-y"], "provider-x")
+      //#when
+      const result = shouldRetryError(error)
 
-    //#then
-    expect(provider).toBe("provider-x")
-  })
+      //#then
+      expect(result).toBe(true)
+    })
 
-  test("treats QuotaExceededError (PascalCase name) as non-retryable STOP error", () => {
-    //#given
-    const error = { name: "QuotaExceededError" }
+    test("triggers fallback for InsufficientCreditsError", () => {
+      //#given
+      const error = { name: "InsufficientCreditsError" }
 
-    //#when
-    const result = shouldRetryError(error)
+      //#when
+      const result = shouldRetryError(error)
 
-    //#then
-    expect(result).toBe(false)
-  })
+      //#then
+      expect(result).toBe(true)
+    })
 
-  test("treats quotaexceedederror (lowercase name) as non-retryable STOP error", () => {
-    //#given
-    const error = { name: "quotaexceedederror" }
+    test("triggers fallback for FreeUsageLimitError", () => {
+      //#given
+      const error = { name: "FreeUsageLimitError" }
 
-    //#when
-    const result = shouldRetryError(error)
+      //#when
+      const result = shouldRetryError(error)
 
-    //#then
-    expect(result).toBe(false)
-  })
+      //#then
+      expect(result).toBe(true)
+    })
 
-  test("treats InsufficientCreditsError (PascalCase name) as non-retryable STOP error", () => {
-    //#given
-    const error = { name: "InsufficientCreditsError" }
+    test("triggers fallback for quota reset message", () => {
+      //#given
+      const error = { message: "quota will reset after 1 hour" }
 
-    //#when
-    const result = shouldRetryError(error)
+      //#when
+      const result = shouldRetryError(error)
 
-    //#then
-    expect(result).toBe(false)
-  })
+      //#then
+      expect(result).toBe(true)
+    })
 
-  test("treats insufficientcreditserror (lowercase name) as non-retryable STOP error", () => {
-    //#given
-    const error = { name: "insufficientcreditserror" }
+    test("triggers fallback for quota exceeded message", () => {
+      //#given
+      const error = { message: "quota exceeded for this billing period" }
 
-    //#when
-    const result = shouldRetryError(error)
+      //#when
+      const result = shouldRetryError(error)
 
-    //#then
-    expect(result).toBe(false)
-  })
+      //#then
+      expect(result).toBe(true)
+    })
 
-  test("treats FreeUsageLimitError (PascalCase name) as non-retryable STOP error", () => {
-    //#given
-    const error = { name: "FreeUsageLimitError" }
+    test("triggers fallback for 'usage limit has been reached' message", () => {
+      //#given
+      const error = { message: "The usage limit has been reached for your account" }
 
-    //#when
-    const result = shouldRetryError(error)
+      //#when
+      const result = shouldRetryError(error)
 
-    //#then
-    expect(result).toBe(false)
-  })
+      //#then
+      expect(result).toBe(true)
+    })
 
-  test("treats freeusagelimiterror (lowercase name) as non-retryable STOP error", () => {
-    //#given
-    const error = { name: "freeusagelimiterror" }
+    test("triggers fallback for insufficient credits message", () => {
+      //#given
+      const error = { message: "insufficient credits to complete this request" }
 
-    //#when
-    const result = shouldRetryError(error)
+      //#when
+      const result = shouldRetryError(error)
 
-    //#then
-    expect(result).toBe(false)
-  })
+      //#then
+      expect(result).toBe(true)
+    })
 
-  test("treats quota reset message as non-retryable STOP error (no error name)", () => {
-    //#given
-    const error = { message: "quota will reset after 1 hour" }
+    test("triggers fallback for subscription quota message", () => {
+      //#given
+      const error = { message: "Subscription quota exceeded. You can continue using free models." }
 
-    //#when
-    const result = shouldRetryError(error)
+      //#when
+      const result = shouldRetryError(error)
 
-    //#then
-    expect(result).toBe(false)
-  })
+      //#then
+      expect(result).toBe(true)
+    })
 
-  test("treats quota exceeded message as non-retryable STOP error (no error name)", () => {
-    //#given
-    const error = { message: "quota exceeded for this billing period" }
+    test("triggers fallback for spending cap message", () => {
+      //#given
+      const error = { message: "Your project has exceeded its monthly spending cap" }
 
-    //#when
-    const result = shouldRetryError(error)
+      //#when
+      const result = shouldRetryError(error)
 
-    //#then
-    expect(result).toBe(false)
-  })
+      //#then
+      expect(result).toBe(true)
+    })
 
-  test("treats usage limit reached message as non-retryable STOP error (no error name)", () => {
-    //#given
-    const error = { message: "usage limit has been reached for your account" }
+    test("returns false for MessageAbortedError (non-retryable)", () => {
+      //#given
+      const error = { name: "MessageAbortedError" }
 
-    //#when
-    const result = shouldRetryError(error)
+      //#when
+      const result = shouldRetryError(error)
 
-    //#then
-    expect(result).toBe(false)
-  })
+      //#then
+      expect(result).toBe(false)
+    })
 
-  test("treats insufficient credits message as non-retryable STOP error (no error name)", () => {
-    //#given
-    const error = { message: "insufficient credits to complete this request" }
+    test("returns false for ContextLengthError (non-retryable)", () => {
+      //#given
+      const error = { name: "ContextLengthError" }
 
-    //#when
-    const result = shouldRetryError(error)
+      //#when
+      const result = shouldRetryError(error)
 
-    //#then
-    expect(result).toBe(false)
-  })
+      //#then
+      expect(result).toBe(false)
+    })
 
-  test("treats 'bad request' message as retryable (GitHub Copilot rolling update)", () => {
-    //#given
-    const error = { message: "400 Bad Request" }
+    test("returns false for ValidationError (non-retryable)", () => {
+      //#given
+      const error = { name: "ValidationError" }
 
-    //#when
-    const result = shouldRetryError(error)
+      //#when
+      const result = shouldRetryError(error)
 
-    //#then
-    expect(result).toBe(true)
+      //#then
+      expect(result).toBe(false)
+    })
+
+    test("returns false for unknown error with no matching pattern", () => {
+      //#given
+      const error = { message: "some completely unrelated error" }
+
+      //#when
+      const result = shouldRetryError(error)
+
+      //#then
+      expect(result).toBe(false)
+    })
   })
 
-  test("treats 'bad request' lowercase as retryable", () => {
-    //#given
-    const error = { message: "bad request: model temporarily unavailable" }
+  describe("#isRetryableModelError", () => {
+    test("returns true for RateLimitError", () => {
+      //#given
+      const error = { name: "RateLimitError" }
+
+      //#when
+      const result = isRetryableModelError(error)
+
+      //#then
+      expect(result).toBe(true)
+    })
+
+    test("returns false for QuotaExceededError (STOP, not retryable on same provider)", () => {
+      //#given
+      const error = { name: "QuotaExceededError" }
+
+      //#when
+      const result = isRetryableModelError(error)
+
+      //#then
+      expect(result).toBe(false)
+    })
 
-    //#when
-    const result = shouldRetryError(error)
+    test("returns false for usage limit message (STOP, not retryable on same provider)", () => {
+      //#given
+      const error = { message: "usage limit has been reached for your account" }
 
-    //#then
-    expect(result).toBe(true)
+      //#when
+      const result = isRetryableModelError(error)
+
+      //#then
+      expect(result).toBe(false)
+    })
+
+    test("returns false for MessageAbortedError (non-retryable)", () => {
+      //#given
+      const error = { name: "MessageAbortedError" }
+
+      //#when
+      const result = isRetryableModelError(error)
+
+      //#then
+      expect(result).toBe(false)
+    })
   })
 
-  test("treats subscription quota message as non-retryable", () => {
-    //#given
-    const error = { message: "Subscription quota exceeded. You can continue using free models." }
+  describe("#isStopModelError", () => {
+    test("returns true for QuotaExceededError", () => {
+      //#given
+      const error = { name: "QuotaExceededError" }
+
+      //#when
+      const result = isStopModelError(error)
+
+      //#then
+      expect(result).toBe(true)
+    })
+
+    test("returns true for quotaexceedederror (lowercase)", () => {
+      //#given
+      const error = { name: "quotaexceedederror" }
+
+      //#when
+      const result = isStopModelError(error)
 
-    //#when
-    const result = shouldRetryError(error)
+      //#then
+      expect(result).toBe(true)
+    })
 
-    //#then
-    expect(result).toBe(false)
+    test("returns true for InsufficientCreditsError", () => {
+      //#given
+      const error = { name: "InsufficientCreditsError" }
+
+      //#when
+      const result = isStopModelError(error)
+
+      //#then
+      expect(result).toBe(true)
+    })
+
+    test("returns true for FreeUsageLimitError", () => {
+      //#given
+      const error = { name: "FreeUsageLimitError" }
+
+      //#when
+      const result = isStopModelError(error)
+
+      //#then
+      expect(result).toBe(true)
+    })
+
+    test("returns true for 'usage limit has been reached' message", () => {
+      //#given
+      const error = { message: "The usage limit has been reached" }
+
+      //#when
+      const result = isStopModelError(error)
+
+      //#then
+      expect(result).toBe(true)
+    })
+
+    test("returns true for quota exceeded message", () => {
+      //#given
+      const error = { message: "quota exceeded for this billing period" }
+
+      //#when
+      const result = isStopModelError(error)
+
+      //#then
+      expect(result).toBe(true)
+    })
+
+    test("returns true for spending cap message", () => {
+      //#given
+      const error = { message: "exceeded its monthly spending cap" }
+
+      //#when
+      const result = isStopModelError(error)
+
+      //#then
+      expect(result).toBe(true)
+    })
+
+    test("returns false for RateLimitError (retryable, not stop)", () => {
+      //#given
+      const error = { name: "RateLimitError" }
+
+      //#when
+      const result = isStopModelError(error)
+
+      //#then
+      expect(result).toBe(false)
+    })
+
+    test("returns false for MessageAbortedError (non-retryable, not stop)", () => {
+      //#given
+      const error = { name: "MessageAbortedError" }
+
+      //#when
+      const result = isStopModelError(error)
+
+      //#then
+      expect(result).toBe(false)
+    })
+
+    test("returns false for unknown error with no stop pattern", () => {
+      //#given
+      const error = { message: "some random error" }
+
+      //#when
+      const result = isStopModelError(error)
+
+      //#then
+      expect(result).toBe(false)
+    })
   })
 
-  test("treats HTTP 429 rate limit message as retryable", () => {
-    //#given
-    const error = { message: "429 Too Many Requests: rate limit reached" }
+  describe("#selectFallbackProvider", () => {
+    test("prefers first connected provider in preference order", () => {
+      //#given
+      readConnectedProvidersCacheSpy?.mockReturnValue(["anthropic", "nvidia"])
+
+      //#when
+      const provider = selectFallbackProvider(["anthropic", "nvidia"], "nvidia")
+
+      //#then
+      expect(provider).toBe("anthropic")
+    })
+
+    test("falls back to next connected provider when first is disconnected", () => {
+      //#given
+      readConnectedProvidersCacheSpy?.mockReturnValue(["nvidia"])
+
+      //#when
+      const provider = selectFallbackProvider(["anthropic", "nvidia"])
+
+      //#then
+      expect(provider).toBe("nvidia")
+    })
+
+    test("uses provider preference order when cache is missing", () => {
+      //#given - no cache file
+
+      //#when
+      const provider = selectFallbackProvider(["anthropic", "nvidia"], "nvidia")
+
+      //#then
+      expect(provider).toBe("anthropic")
+    })
+
+    test("uses connected preferred provider when fallback providers are unavailable", () => {
+      //#given
+      readConnectedProvidersCacheSpy?.mockReturnValue(["provider-x"])
 
-    //#when
-    const result = shouldRetryError(error)
+      //#when
+      const provider = selectFallbackProvider(["provider-y"], "provider-x")
 
-    //#then
-    expect(result).toBe(true)
+      //#then
+      expect(provider).toBe("provider-x")
+    })
   })
 })
 
diff --git a/src/shared/model-error-classifier.ts b/src/shared/model-error-classifier.ts
index b20918d189..3362531992 100644
--- a/src/shared/model-error-classifier.ts
+++ b/src/shared/model-error-classifier.ts
@@ -48,6 +48,8 @@ const RETRYABLE_MESSAGE_PATTERNS = [
   "insufficient",
   "too many requests",
   "over limit",
+  "hit your limit",
+  "hit the limit",
   "overloaded",
   "bad gateway",
   "bad request",
@@ -81,7 +83,6 @@ const RETRYABLE_MESSAGE_PATTERNS = [
 const STOP_MESSAGE_PATTERNS = [
   "quota will reset after",
   "quota exceeded",
-  "usage limit has been reached",
   "free usage limit",
   "billing limit",
   "billing hard limit",
@@ -96,7 +97,9 @@ const STOP_MESSAGE_PATTERNS = [
   "insufficient balance",
   "credit balance",
   "usage limit for this month",
+  "usage limit has been reached",
   "exhausted your capacity",
+  "spending cap",
 ]
 
 const AUTO_RETRY_GATE_PATTERNS = [
@@ -153,11 +156,50 @@ export function isRetryableModelError(error: ErrorInfo): boolean {
 }
 
 /**
- * Determines if an error should trigger a fallback retry.
- * Returns true for deadstop errors that completely halt the action loop.
+ * Determines if an error indicates provider exhaustion (quota/billing/usage limits).
+ * These errors mean the current provider cannot serve requests, but a different provider might.
+ */
+export function isStopModelError(error: ErrorInfo): boolean {
+  if (error.name) {
+    const errorNameLower = error.name.toLowerCase()
+    if (STOP_ERROR_NAMES.has(errorNameLower)) {
+      return true
+    }
+  }
+
+  const msg = error.message?.toLowerCase() ?? ""
+  return STOP_MESSAGE_PATTERNS.some((pattern) => msg.includes(pattern))
+}
+
+/**
+ * Determines if an error should trigger cross-provider fallback.
+ * Returns true for both retryable errors AND stop errors (provider exhaustion),
+ * since a different provider may still be able to serve the request.
+ * Returns false only for non-retryable errors (user aborts, validation, syntax)
+ * where no provider switch would help.
  */
 export function shouldRetryError(error: ErrorInfo): boolean {
-  return isRetryableModelError(error)
+  if (error.name) {
+    const errorNameLower = error.name.toLowerCase()
+    if (NON_RETRYABLE_ERROR_NAMES.has(errorNameLower)) {
+      return false
+    }
+    if (RETRYABLE_ERROR_NAMES.has(errorNameLower) || STOP_ERROR_NAMES.has(errorNameLower)) {
+      return true
+    }
+  }
+
+  const msg = error.message?.toLowerCase() ?? ""
+
+  if (STOP_MESSAGE_PATTERNS.some((pattern) => msg.includes(pattern))) {
+    return true
+  }
+
+  if (hasProviderAutoRetrySignal(msg)) {
+    return true
+  }
+
+  return RETRYABLE_MESSAGE_PATTERNS.some((pattern) => msg.includes(pattern))
 }
 
 /**