21 models

Every model.
One key.

A single OpenAI-compatible endpoint for 21 models. Set spend caps per key, see costs before they land, route with confidence.

Start building freeBrowse models →
ModelProviderInput / 1M tokensTypical p50
Claude Haiku 4.5Anthropic$0.80 / 1M85ms
Claude Sonnet 4.5Anthropic$3.00 / 1M175ms
Claude Opus 4.5Anthropic$15.00 / 1M290ms
Claude 3.7 SonnetAnthropic$3.00 / 1M187ms
GPT-4.1OpenAI$2.00 / 1M195ms
GPT-4.1 MiniOpenAI$0.40 / 1M130ms
View all 21 models →

How we think

Inference infrastructure
should be predictable.

01

Control at every level.

Scoped keys per project. Spend ceilings that hard-stop at the cap. Rate limits you define. Your account is never at the mercy of a single key.

02

Cost you can see coming.

Every streaming response carries a cost estimate header. You know what a request cost before you check your dashboard. No end-of-month surprises.

03

Routing you can trust.

Every response tells you which provider handled the request and why. Fallback chains are logged, not hidden. You own the decision, not the black box.

Platform

Everything a production
integration needs.

OpenAI-compatible

Drop in your Relixr base URL. No SDK changes. Works with LangChain, LlamaIndex, and every OpenAI wrapper.

Intelligent fallback

Provider returns 5xx — Relixr retries the next-best option in under 200ms. You never see the failure.

Cost attribution

Every request tagged to a key, a project, a model. Spend dashboard shows exactly where money went.

Scoped API keys

Keys with model allow-lists, rate limits, spend ceilings. Revoke instantly. Full audit log.

Streaming support

Server-sent events for chat completions with cost estimate headers on every stream. OpenAI-compatible chunk format.

Spend caps

Per-key daily spend limits and workspace balance reserve. Hard-stop at the cap with 402 insufficient_balance — no overrun.

How billing works

Top up.
Use what you need.

Add credits to your account. Pay per token at cost — no markup on model price, just a small platform fee on top.

No subscriptions. No tiers that gate models. Credits never expire. Your balance is reserved before each request — you are never charged more than you hold.

See model pricing →Compare plans →
Model cost
At cost
Passed through wholesale
Platform fee
5%
On top of model cost
Minimum balance
$0
No minimum top-up
Credit expiry
Never
Use at your own pace
Payment
Adapts to you
Local options for United States