all four models · live—openai-sdk compatible · $5 minimum top-up · no subscription
gpt, at the price of gpt.
›↵
$0.375in · $2.25 out / 1m tokens the same prompt, the same sdk, a fraction of the bill. no account required to try.
models · 4 active · usd / 1m tokensupdated may 12, 2026
model
input
cached
output
ctx
best for
gpt-5.4-mini· recommended
$0.375
$0.0375
$2.25
128k
everyday chat, study, bots
gpt-5.3-codex-spark
$0.875
$0.0875
$7.00
128k
coding & developer automations
gpt-5.4
$1.25
$0.125
$7.50
256k
balanced quality
gpt-5.5
$2.50
$0.25
$15.00
256k
demanding, high-stakes prompts
cached input billed at 1/10× when prefixes repeatlong-context tiers apply past threshold on selected models
quickstart · the openai client, pointed elsewhere
01 · INSTALL
# any openai-compatible client works
$ npm install openai
$ export REALROUTER_KEY=sk_rr_•••
02 · CALL
import OpenAI from "openai"
const client = new OpenAI({
baseURL: "https://api.realrouter.com/v1",
apiKey: process.env.REALROUTER_KEY,
})
const r = await client.responses.create({
model: "gpt-5.4-mini",
input: "plan my week",
})
↳ same shapes for chat.completions, responses, streaming, tools, vision.full docs → capabilities · what the api supports
✓streaming
semantic sse events, on by default
✓function calling
tools, structured outputs
✓vision
image inputs on supported models
✓web search
fresh context on demand
✓cached input
1/10× on repeated prefixes
✓mobile
companion app, same balance
may 8
models
gpt-5.5 added to the catalogue
→
may 4
platform
companion mobile app · same key, same balance
→
apr 30
billing
cached input billing live · repeated prefixes at 1/10 cost
→
apr 22
api
streaming defaulted to semantic sse events
→
apr 10
billing
starter $5 top-up live · stripe checkout
→