ai-seller.vibe-codes.ru/v1 · OpenAI-compatible
Every top model.
One key.
One key for GPT, Claude, Gemini, Grok, images and video — at wholesale pricing, up to 80% below direct rates. Limits, analytics and caching are already built in.
- −80%
- below direct rates on top models
- 60+
- models: text, images, video
- 99.98%
- gateway uptime over 90 days
One key — any model answers
only "model" changesPoint base_url at our address — in the OpenAI SDK, Claude Code, Cursor or plain curl. After that you only change model: the whole catalogue runs off one balance — text, images and video. The router takes each request to the right provider itself.
The other models stay in orbit around the same key — switching takes no code changes.
active route · provider: anthropic
Wholesale pricing
$ per 1M tokens · output- gpt-5.2$4.40
$15.70−72% - claude-sonnet-4.5$6.00
$15.00−60% - gemini-3-pro$3.60
$10.00−64% - veo-3.1-fast$0.087per second of video−56%
Plug in your volume — honest numbers, no “up to”
Calculated at an average 50/50 input–output split, caching not counted. With context caching, agent sessions get up to 87% cheaper still.
full catalogue of 60+ modelsSavings calculator
- direct
$491- via ai-seller
- $138
- saved / month
- $354
Context caching
0.1× of the price
Repeated prompts and conversation history are read from cache — a typical agent session costs 87% less.
Uptime over 90 days
99.98%
Direct routes to providers. Status and success ratio for each one live in analytics, so degradations show up immediately.
Pay for what you use
$0 in subscriptions
Only the tokens, images and seconds of video you actually spent. Top up, spend, see it in analytics.
One invoice for the whole team
keys · analytics · balanceKeys with spend limits
Every key gets its own ceiling in USD. Rotation, instant revoke, anonymous keys for contractors.
Token analytics
Spend, latency and success ratio by day, model and key. With filters, tooltips and CSV export.
Shared balance
One account for everyone: card, crypto or a company invoice with closing documents. Auto top-up on a threshold — your agents will not stall overnight.
From sign-up to your first request — three minutes.
$5 to start · about 2M gpt-5.2-mini tokens
Frequently asked questions
faqAre these the same models as going direct to the providers?
Yes. Requests go to the official OpenAI, Anthropic, Google and xAI APIs — same weights, same quality, current versions. Only the price changes — and the fact that the whole catalogue sits behind one key.
Why does it come out up to 80% cheaper?
We buy tokens wholesale at the volume of the entire platform and pass the wholesale price on. Plus context caching: repeated prompts and conversation history are billed at 0.1× the price — a typical agent session gets up to 87% cheaper still.
How do I connect? Is it OpenAI SDK compatible?
Point base_url at ai-seller.vibe-codes.ru/v1, drop in your key — that is it. Works with the OpenAI SDK, LangChain, Cursor, Claude Code and plain curl, with no code rewrites.
How do I top up, and is there a subscription?
There is no subscription — you pay only for the tokens, images and seconds of video you used. Top up by card or crypto; companies get an invoice and closing documents. Start from $5.
Do you store the contents of my requests?
Request and response bodies are not stored. For billing and analytics we keep metadata only: model, token counts, latency and response status.
What are the rate limits?
The rate limit is shared platform-wide and sits above providers' entry tiers. Spend ceilings per key you set yourself — in USD per month, with instant revoke.