The LLM Fallback Chain Picker

# The LLM Fallback Chain Picker
# A free tool from VernaOne — https://verna.one/fallback-chain-picker
# Paste this whole message into ChatGPT, Claude, Gemini, or any capable LLM,
# then fill in the three lines at the bottom. No signup, no install.
# ---------------------------------------------------------------------------

You are a senior AI infrastructure engineer. Design a resilient, cost-aware
fallback chain for the task below: an ordered list of LLMs (from different
providers) that produce equivalent-quality results, so the application keeps
working — and stays cheap — when the primary model is down, rate-limited,
deprecated, or too expensive.

Return exactly these sections, in Markdown. Be specific and terse. No preamble.

## 1. Recommended chain
An ordered list of 3–4 models: **Primary → Fallback 1 → Fallback 2 (→ 3)**.
Rules:
- each fallback must be from a **different provider** than the one before it,
  so a single provider outage never takes you fully down;
- order by best **quality-per-dollar** for THIS task, not raw capability;
- note the approx input/output price (per 1M tokens) next to each.

## 2. Why each one
One line per model: what it's good for here, and the specific failure it covers
(outage, rate limit, price spike, context-length overflow, deprecation).

## 3. Routing rules
- when to fail over (timeout, HTTP 429/5xx, refusal, empty output);
- any per-model prompt tweaks needed so output stays consistent across the chain
  (keep these to a minimum — prefer a model-agnostic prompt);
- one cheap "last resort" model that will at least return *something* usable.

## 4. Ready-to-use config
A JSON block like:
{ "primary": "...", "fallbacks": ["...", "..."], "failOn": ["timeout","429","5xx"] }

---
Fill these in:
TASK: {{what the prompt does, e.g. "summarize support tickets into JSON"}}
QUALITY BAR: {{e.g. "must follow the JSON schema exactly; tone doesn't matter"}}
PRIORITY: {{"cheapest that works" | "fastest" | "highest quality within budget"}}