The simplest LLM API: don't pick a model, pick a tier

You want to add AI to your app. You don't want to become a model sommelier first. But every path in makes you choose from dozens of checkpoints with version-soup names, compare per-token price tables that change monthly, and re-run that research each time a provider ships or deprecates something. The comparison sites answer "which model is best?" with a longer list of models.

If your actual requirement is "cheap and fast," "balanced," or "smartest available" — you're choosing a performance class, not a model. TierUp lets you say exactly that and nothing more: one OpenAI-compatible endpoint, four tiers, no model names anywhere in your code.

The whole integration

curl https://tierup.ai/v1/chat/completions \
  -H "Authorization: Bearer $TIERUP_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "tier-2",
    "messages": [{"role": "user", "content": "Draft a reply to this email."}]
  }'

If you already use an OpenAI SDK, switching is two strings: the base URL and the model field. Streaming, error shapes, and usage accounting follow the conventions your client already expects. We map the tier to what we assess as the best-value model in that class server-side and forward the call; when a better model enters the class, the tier upgrades with zero changes on your end.

"Which tier should I use?" is a question you can answer

Unlike "which model should I use?", this one doesn't require reading benchmarks — you already know what your feature needs:

Prices per 1 million tokens. Wrong guess? Changing tiers is editing one string, so start cheap and move up when an eval tells you to.

What it costs, honestly

Per-tier prices run ~50% below the retail price of the models we route to, and tier-1 is currently free (capped at 100 requests/day until your first top-up). The caveat we put everywhere, including here: that discount is a transparent subsidy during our public beta — capped by a daily spend guardrail, not secret efficiency — and we route through OpenRouter today, so this is an abstraction on top of existing infrastructure, not our own GPU fleet. Billing is prepaid credits: $25 free on signup, no card required.

Doesn't "we pick for you" mean it changes under me?

The model behind a tier can change — that's the work you're delegating. What we commit to is that it never changes silently: every model/price/free change to a named tier is versioned, material changes are announced at least 14 days ahead, the record is machine-checkable via GET /v1/tiers/changelog, and CI physically refuses to let a tier drift without a changelog entry. A named tier is a stable contract, not a moving target — details on the Tier Changelog.

When "simplest" is the wrong goal

• You need to pin an exact model for reproducibility, compliance, or benchmarking — use a marketplace like OpenRouter or the provider directly. We strip model names from responses on purpose.
• You want routing rules you control — fallback chains, per-model observability, self-hosting — that's a configured LLM router or gateway, a different tool.
• You need a mature vendor in the serving path today. We're early and honest about it: built solo, public beta, ~zero production users.

FAQ

Also see: How tier routing works · Tier pricing & details · LLM API cost calculator · Cheapest AI API in 2026 · Tier Changelog