An OpenAI-compatible gateway at a fraction of the vendors' own list prices. Change one line in your SDK and keep everything else — no VPN, no subscription, pay in rubles, USDT or by card.
Same models, same ids, same API shapes. The savings come from how the capacity is bought, not from what you receive.
response.model always echoes the model you asked for. Nothing is quietly downgraded to something cheaper — check it on any call.
Anthropic charges $10 per 1,000 web_search calls on its own API; Google bills grounded requests separately. Here search results bill as ordinary input tokens.
The gateway handles the upstream side. Register, top up from $5, make a request.
SBP, Russian cards and SberPay; USDT across four networks; BTC, ETH, SOL and more; card and PayPal internationally.
USD per million tokens, against each vendor's official list price. Savings are blended 91% input / 9% output — roughly how a coding-agent workload splits.
| Model | Input / 1M | Output / 1M | Official | You save |
|---|---|---|---|---|
| Claude | ||||
| claude-fable-5 | $4.42 | $22.12 | $10.00 | −56% |
| claude-opus-4-7 | $0.667 | $3.33 | $5.00 | −87% |
| claude-opus-4-6 | $0.66 | $3.30 | $5.00 | −87% |
| claude-opus-5 | $0.66 | $3.33 | $5.00 | −87% |
| claude-opus-4-8 | $0.66 | $3.30 | $5.00 | −87% |
| claude-sonnet-4-6 | $0.4 | $2.00 | $3.00 | −87% |
| claude-haiku-4-5-20251001 | $0.133 | $0.667 | $1.00 | −87% |
| GPT & Codex | ||||
| gpt-5.5 | $0.74 | $4.43 | $5.00 | −85% |
| gpt-5.5-openai-compact | $0.74 | $5.90 | $5.00 | −83% |
| gpt-5.6-sol | $0.74 | $4.36 | $5.00 | −85% |
| gpt-5.6-terra | $0.3 | $1.77 | $2.50 | −88% |
| gpt-5.6-luna | $0.14 | $0.84 | $1.00 | −86% |
| Gemini | ||||
| gemini-3.1-pro-low | $0.467 | $2.80 | $2.00 | −77% |
| gemini-3.1-pro-preview | $0.467 | $2.80 | $2.00 | −77% |
| gemini-3.5-flash | $0.35 | $2.10 | $1.50 | −77% |
| gemini-3-flash | $0.117 | $0.7 | $0.5 | −77% |
| Grok | ||||
| grok-4.5 | $0.65 | $2.00 | $2.00 | −67% |
Image and per-request models are on the full price page. The live catalog also shows cache-read and cache-write rates.
The official OpenAI SDK, pointed somewhere else. Nothing else in your code changes.
from openai import OpenAI client = OpenAI( api_key="YOUR_API_KEY", base_url="https://cheapai.io/v1", ) r = client.chat.completions.create( model="claude-opus-4-8", messages=[{"role": "user", "content": "Hello"}], ) print(r.choices[0].message.content)
curl https://cheapai.io/v1/chat/completions \ -H "Authorization: Bearer $CHEAPAI_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "model": "claude-opus-4-8", "messages": [{"role":"user","content":"Hello"}] }'
https://cheapai.io with no /v1 — the SDK appends /v1/messages itself.More runnable samples — streaming, vision, tool calling, prompt caching — in the examples repository.
Anything that accepts a custom endpoint. Each guide carries the quirk that silently breaks that particular tool.
Base URL takes no /v1 — the one that catches everyone.
A leftover OpenAI login silently wins over your config.
Pro required — the Free plan blocks custom keys outright.
The key lives in the keychain, not in settings.json.
There is no custom provider — you write openai.
Same OpenAI-compatible provider, three different settings panels.
The rest are on the FAQ page.
Yes — same model ids, same API shapes, same capabilities. response.model echoes exactly what you requested, so you can verify it on any call. Nothing is silently swapped for a cheaper model.
Volume purchasing upstream plus a thin margin on top. There is no subscription and no monthly minimum — a prepaid balance is drawn down per token.
No. The gateway handles the upstream side, so it works from wherever you are.
Claude (Opus, Sonnet, Haiku, Fable), the GPT-5.x line including Codex variants, Gemini 3.x Pro and Flash, Grok 4.x, plus image models billed per request. The live catalog is always current.
Yes, on Claude models, billed at the reduced cache-read rate. There is a runnable example.
Pay in rubles, USDT or by card. No VPN, no subscription, from $5.
Get an API key