{
  "title": "Model fallback",
  "description": "Switch to another model after a capacity, policy, or service failure.",
  "slug": "model-fallback",
  "locale": "en",
  "seoTitle": "Model fallback compatibility — Can My Agent Use",
  "socialTitle": "Model fallback",
  "socialDescription": "Compare documented fallback routing, disclosure, configuration, and retry behavior.",
  "llmSummary": "Model fallback is documented switching to another model after a capacity, policy, or service failure. Disclosure, configuration, and opt-out behavior are recorded as qualifiers.",
  "audience": "Engineers who need predictable model behavior and reproducible runs.",
  "contentKind": "feature",
  "status": "published",
  "tags": [
    "models",
    "routing",
    "reliability",
    "observability"
  ],
  "updated": "2026-08-28T00:00:00.000Z",
  "published": "2026-08-28T00:00:00.000Z",
  "category": "models-context",
  "summary": "Switch to another model after a documented failure condition.",
  "specLabel": "Common product term",
  "aliases": [
    "model failover",
    "automatic model routing",
    "fallback model"
  ],
  "capabilityKind": "atomic",
  "parent": "models-and-context",
  "related": [
    "model-selection",
    "resumable-runs"
  ],
  "relations": [],
  "highlight": false,
  "notes": [
    {
      "id": 1,
      "text": "Evidence checked 2026-08-28: xAI documents automatic model failover for Grok Bot and says usage analytics show the model that actually served each request, including failovers. Users and administrators cannot choose models or disable fallback."
    },
    {
      "id": 2,
      "text": "Evidence checked 2026-08-28: Gemini CLI enables model fallback by default after quota or server errors, normally prompts before switching, and can apply policy-authorized silent fallback for a turn or session. Some internal utility calls use a separate silent fallback chain."
    }
  ],
  "issues": [],
  "resources": [
    {
      "title": "Methodology",
      "href": "/methodology",
      "kind": "note"
    },
    {
      "id": "xai-grok-bot-team-models",
      "title": "xAI — Grok Bot for teams and enterprises",
      "href": "https://docs.x.ai/grok-bot/teams-and-enterprises",
      "kind": "docs",
      "publisher": "xAI",
      "evidenceType": "documented",
      "reviewedAt": "2026-08-28",
      "locator": "Models and providers"
    },
    {
      "id": "google-gemini-cli-model-routing",
      "title": "Google — Gemini CLI model routing",
      "href": "https://geminicli.com/docs/cli/model-routing/",
      "kind": "docs",
      "publisher": "Google",
      "evidenceType": "documented",
      "reviewedAt": "2026-08-28",
      "locator": "How it works"
    }
  ],
  "support": [
    {
      "harness": "grok-bot-desktop",
      "versions": [
        {
          "track": "current",
          "status": "yes",
          "noteIds": [
            1
          ],
          "target": {
            "kind": "hosted-observation",
            "revision": "2026-08-28 Grok Bot desktop documentation observation",
            "observedAt": "2026-08-28"
          },
          "environmentProfile": "hosted-default",
          "qualifiers": [
            {
              "type": "runtime",
              "value": "automatic failover is documented and the serving model appears in usage analytics, but the operator cannot configure, select, or prevent fallback"
            }
          ],
          "evidence": [
            {
              "resourceId": "xai-grok-bot-team-models",
              "type": "documented",
              "observedAt": "2026-08-28"
            }
          ]
        }
      ]
    },
    {
      "harness": "gemini-cli",
      "versions": [
        {
          "track": "current",
          "status": "yes",
          "noteIds": [
            2
          ],
          "target": {
            "kind": "hosted-observation",
            "revision": "2026-08-28 Gemini CLI documentation observation",
            "observedAt": "2026-08-28"
          },
          "environmentProfile": "local-default",
          "qualifiers": [
            {
              "type": "runtime",
              "value": "model fallback is enabled by default and can trigger after quota or server errors; the CLI normally prompts for consent before switching"
            },
            {
              "type": "policy",
              "value": "a routing policy can permit silent fallback for the current turn or remainder of the session; internal utility calls have a separate silent fallback chain"
            },
            {
              "type": "runtime",
              "value": "the reviewed page identifies fallback behavior but does not provide an exported per-turn routing provenance record"
            }
          ],
          "evidence": [
            {
              "resourceId": "google-gemini-cli-model-routing",
              "type": "documented",
              "observedAt": "2026-08-28"
            }
          ]
        }
      ]
    }
  ],
  "html": "/features/model-fallback",
  "markdown": "/features/model-fallback.md",
  "json": "/api/v1/features/model-fallback.json",
  "body": "This row tracks whether a harness may retry or continue with a different model and what control the operator has over that change. Transparent, configurable fallback is distinct from silent automatic routing.\n\nEvidence should record triggers, candidate models, changes in context or tool compatibility, whether the operator is notified, how provenance appears in logs or exports, and whether regulated or reproducible workflows can disable fallback."
}