Relay
relay online · gateway live · 8 concurrent / account

Frontier models.
Up to 60% below list.

Relay resells SearchAtlas AI Gateway capacity at per-model prices, up to 60% below the gateway’s list price — one OpenAI- and Anthropic-compatible endpoint, every frontier model, metered to the request. No app-code rewrite.

Fund $5, get $15 — +$10 on your first deposit at paid launch
quickstart.py
from openai import OpenAIclient = OpenAI(    base_url="https://ai-gateway.getborg.com/api/v1/proxy/v1",    api_key="$RELAY_API_KEY",)resp = client.chat.completions.create(    model="glm-5.3",    messages=[{"role": "user", "content": "Summarize this ticket"}],)

Pricing

Same models. 93 of 100 priced below list.

List is the gateway's own list price. The Relay price is what each request is debited.

Paid launch is not open yet: card payments are not live.

ModelList / 1M outRelay / 1M outDiscount
glm-5.3$4.40$2.42-45%
glm-5.2$4.40$2.42-45%
glm-5-turbo$4.00$2.40-40%
google/gemini-3.7-flash$3.75$2.44-35%
google/gemini-3.6-flash$3.75$2.44-35%
google/gemini-3.1-pro-preview$12.00$10.11-16%

See every model and both input + output prices →

Three steps

Sign up, fund, swap the base URL

  1. 1

    Sign up

    Create an account and verify with a card — no KYC paperwork, no sales call.

  2. 2

    Fund your balance

    Deposit $5 or more. At paid launch, a first deposit is planned to get a flat +$10 bonus (not live yet).

  3. 3

    Swap the base URL

    Point your existing OpenAI or Anthropic SDK at the Relay endpoint. Everything else stays the same.

Drop-in

Speaks OpenAI and Anthropic, natively

Keep your SDK, your tooling, your coding agent. Only the base URL and key change.

OpenAI-compatible

Chat Completions and Responses — same request/response shape as api.openai.com.

base_url = https://ai-gateway.getborg.com/api/v1/proxy/v1

Anthropic-compatible

The Messages API, unchanged — point the Anthropic SDK or Claude Code at Relay with one env var.

base_url = https://ai-gateway.getborg.com/api/v1/proxy

Terms

No surprises in the fine print

No prompt storage

We do not store prompt or response bodies. Metadata only — model, tokens, cost.

Pay as you go

Prepaid credits, debited per request at the rates shown in the model catalog. No seats, no usage tiers.

No monthly fee

Credits don't expire while your account is active. Fund once, use over time.

Cancel any time

No contract. Stop sending requests whenever you want — nothing to unwind.

Lower inference costs are coming soon

Paid launch is pending. At launch, a $5 minimum first deposit qualifies for a flat $10 bonus.