NixAPI中Console
LatestGroup launch: jev-1, hunyuan-1; Model launch: jev-1.13.0, hy4-preview

One API.
Every model.
Routed.

An OpenAI-compatible gateway to 370+ LLMs — automatic failover, one bill, and prices below official rates.

Get your API key →Read the docs
nixapi · routerLIVE
POST /v1/chat/completionsAuthorization: sk-…Your appOpenAI SDKNixAPIsmart routingClaude286msGPT231msGemini412msDeepSeek198msQwen174msGrok344msYour appOpenAI SDKNixAPIsmart routingClaudeGPTGemini+ DeepSeek, Qwen, Kimi… 370+ models in total

ONE KEY FOR MODELS FROM

OpenAIAnthropicGoogleDeepSeekxAIMidjourney
QwenMoonshotZhipuDoubaoMiniMaxKling

Send one request. Routed to the best upstream. Failover in milliseconds. One bill for everything.

01 — Request

Send one request.

Keep the OpenAI SDK you already use. Point base_url at NixAPI and every model sits behind a single key.

02 — Route

Routed to the best upstream.

Each call goes to the healthiest, fastest channel for that model, measured continuously.

03 — Failover

Failover in milliseconds.

When a provider slows down or errors out, traffic moves to a healthy route before your users notice.

04 — Billing

One bill for everything.

Tokens from every provider land on a single, itemised balance. Top up once, spend on any model.

your appPOST /v1/chatNixAPIrouterClaude286msGPT231mstimeoutGemini412msDeepSeek198ms1 key✓ fastest healthy✕ 503 upstream↻ rerouted · 12msBALANCE · this month$1,284.604 providers · 1 invoice
Nix/nɪks/verb — to get rid of; to strike out.

Ten providers. Ten API keys. Ten invoices. Ten ways to fail at 3 a.m. NixAPI turns all of it into one endpoint that just works.

012345678901234567890123456789,012345678901234567890123456789,012345678901234567890123456789,012345678901234567890123456789
tokens processed
372models online
24model providers
−85%avg. savings on popular models
Why NixAPI

Built for teams that ship on AI.

Everything you need to run production LLM traffic — without juggling accounts, SDKs and invoices.

Drop-in OpenAI compatible

Works with the official SDKs, LangChain, LlamaIndex, Dify, Cursor and any client that speaks the OpenAI format.

base_url = "https://api.nixapi.com/v1"

Pay as you go

No subscription, no minimum. Top up and spend on any model.

$0

monthly minimum

Low-overhead relay

A lightweight gateway talks to upstreams directly, adding only tens of milliseconds.

Keys, limits and logs

Create keys per project, cap spend, restrict models and inspect every request from the console.

sk-7f3a••••••••2c9e● active · $50 cap

Human support

Something looks off? Chat with a real engineer from the bottom-right corner, or drop us an email.

Live chatEmail
Reliability

High availability, by default.

Active-active redundancy and round-the-clock operations keep every call landing.

99.95%Uptime, last 30 days (per model)
32msAvg. gateway latency
No capsHigh concurrency, no throttling or queueing at peak
  1. 01

    Active-active redundancy

    Core services run redundantly across nodes and fail over within seconds.

  2. 02

    Smart scheduling

    Upstream channels are probed continuously; requests take the fastest healthy path.

  3. 03

    24/7 operations

    An on-call team monitors and responds around the clock, before issues reach you.

Models & pricing

Every frontier model, below list price.

GPT, Claude, Gemini, DeepSeek, Qwen and more — 370+ models behind one endpoint and one balance. Pay only for the tokens you use.

Input
$0.6214
Output
$3.107
Official
$10 / $50
save 94%
Input
$0.3107
Output
$1.864
Official
$5 / $30
save 94%

claude-opus-5

Anthropic
Input
$0.7456
Output
$3.728
Official
$5 / $25
save 85%
Input
$1.115
Output
$3.346
Official
$1.32 / $3.96
save 16%

grok-4.6

Grok (xAI)
Input
$0.2485
Output
$0.7456
Official
$2 / $6
save 88%
All 370+ models →See all models & prices

USD per 1M tokens · Official list price vs. the lowest NixAPI group price at build time. Live rates on the pricing page.

Direct to providers

One endpoint. Every model, worldwide.

Whether a model comes from the US or China, you call it through the same base_url. The gateway keeps checking upstream channels and routes around failing ones.

  • OpenAI · Anthropic · Google · xAI US
  • DeepSeek · Qwen · Kimi · Zhipu China
  • Doubao · MiniMax · Kling China
Security & compliance

Your keys and data stay yours.

Guardrails at every layer: access, transport and audit.

Access control

Group API keys by project, then add IP allowlists, rate limiting, circuit breaking and authentication.

  • Key groups
  • IP allowlist
  • Rate limits
  • Auth

Data privacy

Encrypted in transit end to end. GDPR data processing agreement; request and response bodies are not stored by default.

  • Encrypted
  • GDPR DPA
  • No retention

Audit trail

Key operations are logged and traceable, backed by regular security audits and vulnerability scans.

  • Operation logs
  • Security audits
  • Vuln scans
Quickstart

Change one line. Keep everything else.

  1. Create an accountSign up in seconds — no card required.
  2. Generate a keyPer-project keys with spend limits.
  3. Swap the base URLYour SDK, request body and responses stay the same.
# pip install openaifrom openai import OpenAIclient = OpenAI(-    base_url="https://api.openai.com/v1",+    base_url="https://api.nixapi.com/v1",     api_key="sk-…",)resp = client.chat.completions.create(     model="claude-sonnet-5",     messages=[{"role": "user", "content": "Hello"}],)
Why NixAPI

Five reasons, at a glance.

  1. 01

    Compliant

    GDPR DPA, no call data retained by default, enterprise audit support.

  2. 02

    Reliable

    Three years in production, with load balancing and automatic retries.

  3. 03

    Transparent

    Every model price and group multiplier is public. No hidden fees.

  4. 04

    Compatible

    One base_url for the apps and frameworks that speak the OpenAI protocol.

  5. 05

    Supported

    24/7 technical support, in multiple languages.

FAQ

Questions, answered.

What is NixAPI?

NixAPI is an AI API gateway. It gives you one OpenAI-compatible endpoint and one API key to call models from OpenAI, Anthropic, Google, DeepSeek, Alibaba Qwen and other providers.

Why is it called NixAPI?

"Nix" means to get rid of something. NixAPI nixes the overhead of working with many AI providers (separate keys, SDKs, invoices and markups), so calling any model comes down to one endpoint. The name also nods to *nix (Unix-like) systems: stable, low-level infrastructure built for developers.

Do I need to change my code?

No. Keep your existing OpenAI SDK or HTTP client, change base_url to https://api.nixapi.com/v1 and use your NixAPI key. Request and response formats stay the same.

Which models are supported?

More than 370+ models, including GPT, Claude, Gemini, DeepSeek, Qwen, Kimi, GLM and Grok families. The full, up-to-date list with prices is on the pricing page.

How is usage billed?

Pay as you go. You top up a balance and each request is charged per token at the listed model price. There is no subscription or monthly minimum.

What happens when a provider goes down?

NixAPI health-checks every upstream channel. If one slows down or returns errors, requests are automatically routed to a healthy channel for the same model.

How do I get support?

Reach the team through the contact page in the console. Technical guides and API references are in the documentation.

Start routing.

Create an account, grab a key, and call any model in under a minute.

Get started free →Talk to us