The OpenAI-compatible endpoint

Base URL

https://1xai.ir/v1

Auth

Authorization: Bearer 1xai-…on every request. Keys are minted in the dashboard; a key can also read its own account's usage.

Routes

/v1/chat/completions, /v1/embeddings, /v1/images/generations, /v1/audio/speech, /v1/audio/transcriptions, /v1/moderations, /v1/files, /v1/models.

Native routes

/anthropic/v1/* (Messages API, x-api-key + anthropic-version) and /gemini/v1beta/* (generateContent, x-goog-api-key or ?key=).

Streaming

"stream": true returns Server-Sent Events; stream_options.include_usage is added for you, so the terminal chunk carries token counts.

Models

Over 38, across OpenAI, Anthropic, Google and DeepSeek. The live list with Toman prices is on the models page.

The reason to use the compatibility route rather than four SDKs: the request body does not change when the provider does. The same call reaches a Google model here, and an Anthropic or OpenAI one by editing a single field — which makes model A/B tests, per-tenant model choice and graceful fallback ordinary code rather than an integration project.

// npm i openai
import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://1xai.ir/v1",
  apiKey: process.env.ONEXAI_API_KEY, // "1xai-..."
});

const stream = await client.chat.completions.create({
  model: "gemini-2.5-flash",   // or gpt-4o-mini, claude-haiku-4-5, deepseek-chat
  messages: [{ role: "user", content: "List three uses for an OCR pipeline." }],
  stream: true,
});

for await (const chunk of stream) {
  process.stdout.write(chunk.choices[0]?.delta?.content ?? "");
}

Chat is the busiest route but not the only one. Embeddings, image generation, speech synthesis and transcription all sit under the same base URL with the same key and the same wallet, so a pipeline that mixes them needs one credential.

# every model your key can address, OpenAI /v1/models shape
curl https://1xai.ir/v1/models \
  -H "Authorization: Bearer 1xai-..."

# embeddings, same base URL, same key
curl https://1xai.ir/v1/embeddings \
  -H "Authorization: Bearer 1xai-..." \
  -H "Content-Type: application/json" \
  -d '{"model": "text-embedding-3-small", "input": "سلام دنیا"}'

Cost visibility is part of the API rather than a dashboard-only feature. The same 1xai-key reads the account's call log and daily roll-ups, which is what you want when you are attributing spend to a tenant or wiring a budget alarm.

# what you spent, per call — no dashboard scraping, no second token
curl "https://1xai.ir/v1/usage?limit=50&status=success&model=gpt-4o-mini" \
  -H "Authorization: Bearer 1xai-..."

# daily roll-up for the last 30 days
curl "https://1xai.ir/v1/usage/summary?days=30" \
  -H "Authorization: Bearer 1xai-..."

A single wire format is a simplification, and simplifications lose things. Anthropic's cache_control, extended thinking, citations and server-side tools have no field in the OpenAI schema; neither do Google's cachedContents or grounding. On /v1 those parameters are dropped rather than translated, silently, because there is nothing to translate them into. If your prompt repeats a large fixed block on every call, that is exactly the feature you want, and the native route is where it lives: see the Claude page for a working example.

1xAi is an independent gateway. It is not affiliated with, endorsed by, or a reseller for OpenAI, Anthropic, Google or DeepSeek. Requests are relayed to those providers' official APIs, so each provider's own usage policies apply to what their models will and will not do, and their model availability is theirs to change. Current reachability is published on the status page.