CI/CD & Automation Simple

Fallback Model on Overload

Automatically switch to a cheaper model when the primary model is rate-limited

Command

$ "color:#7C5CFC">claude -p "Review this PR" \
    "color:#d97757">--fallback-model "color:#7C5CFC">claude-haiku-4-5-20251001 \
    "color:#d97757">--max-budget-usd 0.50 \
    "color:#d97757">--output-format json

Response

{
  "subtype": "success",
  "result": "FALLBACK_TEST_DEFAULT",
  "total_cost_usd": 0.01679325,
  "modelUsage": {
    "claude-opus-4-6": {
      "inputTokens": 3,
      "outputTokens": 10,
      "cacheReadInputTokens": 14269,
      "costUSD": 0.01679325
    }
  }
}

Parsing Code

059669">">const data = JSON.parse(output);
059669">">// Check which model was actually used
059669">">const models = Object.keys(data.modelUsage);
console.log(059669">`Used model: ${models[0]}`);

Gotchas

! Fallback ONLY triggers on HTTP 429 rate limit — not on other errors (500, timeout, auth)

! Check modelUsage keys to see which model actually handled the request

Command

Response

Parsing Code

Gotchas

Related Recipes