We test the models
so you don't have to.
One key. Every frontier model. Drop into the tools you already use — Cursor, Claude Code, Cline, or any OpenAI/Anthropic SDK.
100 credits free · No card · 30-second setup
# Anthropic-compatible
curl https://api.omnizen.ai/v1/messages \
-H "x-api-key: $OMNIZEN_KEY" \
-H "anthropic-version: 2023-06-01" \
-d '{
"model": "deepseek-v4",
"max_tokens": 1024,
"messages": [{ "role": "user", "content": "Hello" }]
}'The frontier of open models, behind one URL.
We curate fast, capable, and cost-effective models from the labs shipping today. We add new ones, retire stale ones, and route around outages — so you don't have to.
Browse all models →From sign-up to first response in under a minute.
- 01
Sign up
Email, GitHub, or Google. 100 credits free — about 3M tokens of Smart Quick routing.
→ omnizen.ai/sign-up - 02
Copy your key
One default key, or create scoped keys per project with spend caps.
sk-omz-•••••••••••••••• - 03
Paste into your tool
Cursor, Claude Code, Cline, OpenCode, or any OpenAI/Anthropic SDK. No code rewrite.
ANTHROPIC_BASE_URL=https://api.omnizen.ai
Built for the way developers actually work.
Works with the tools you already pay for.
Cursor, Claude Code, Cline, OpenCode, Aider, Continue. If it speaks OpenAI or Anthropic, it speaks Omnizen.
Pay how you like.
USD card, bKash, or EPS. Priced in USD — no FX surprises.
Spend exactly what you want.
Hard caps per key, auto top-up thresholds, and credits that never expire.
Anthropic and OpenAI shapes, same key.
Use messages.create from one app and chat.completions.create from another. Same backing models, same billing, same logs.
ai.messages.create({ model: 'glm-4.6', ... })
ai.chat.completions.create({ model: 'glm-4.6', ... })
One subscription. Every model.
Starter
Up to 31M tokens / mo
Chat, drafts, light code
Pro
PopularUp to 102M tokens / mo
Daily work and prototypes
Power
Up to 306M tokens / mo
Agents and serious lifting
Questions, answered.
How is this different from ChatGPT, Claude, or Cursor?
We don't lock you to one model. We benchmark every frontier model on common tasks and route your queries to the one that wins. New models ship every 4 weeks; we add them within 14 days. You get the current best model at $5–$40/mo while their subscriptions slowly decay.
What's the difference between Smart Quick, Smart Expert, and Manual?
Smart Quick (default) routes most traffic to fast cheap models like Qwen-Flash and DeepSeek V4-Flash — best for chat, drafts, and light code. Smart Expert routes to quality-first models like Kimi K2.5 and GLM 4.6 — best for reasoning and serious work. Manual lets you pick a specific model when you know what you want. Switch any time.
How does the weekly credit refresh work?
Your monthly credits drip in weekly (Pro = 1,000 credits every Monday). Unused credits roll over up to two weeks' worth, so a vacation doesn't waste anything — but you can't burn through a month's allotment on day one and run dry.
Do you really test every model? Where's the proof?
Yes. We run 20 task categories (chat, code, JSON-strict, long-context, reasoning, translation, agent loops, more) on every frontier release. Public leaderboard is at omnizen.ai/benchmarks. Each routing-table change shows up in the changelog with the score delta.
Can I really use this with Claude Code and Cursor?
Yes. Set ANTHROPIC_BASE_URL or OPENAI_BASE_URL to api.omnizen.ai and your Omnizen key. Tools that call the Anthropic Messages API or OpenAI Chat Completions API will route through us — no code changes.
Do credits expire?
Subscription credits refresh weekly and bank up to 2× the weekly drip. Topup-pack credits (one-shot purchases) never expire — they live in a separate wallet and stack on top of any plan.
What about rate limits?
Cheap models (Qwen-Flash, DeepSeek V4-Flash) are effectively unlimited within an RPM cap that scales by tier (Starter 60, Pro 120, Power 240, Team 600 pooled). Premium models are gated by credits, not time windows.
What about Claude or GPT-4?
BYOK pass-through is available on Pro and above with a 5% gateway fee — bring your Anthropic or OpenAI key, we route through it. They're not bundled in the flat tiers because their list-price economics would force the tiers to $20+.
What data do you log?
Request metadata for billing and abuse prevention plus the routing decision (which model we picked and why). Full prompt/completion logs are off by default — opt in per key from the dashboard for debugging.
Refunds?
Topup-wallet credits are refundable on request within 30 days. Subscription fees are non-refundable mid-cycle but you can downgrade or cancel any time.
Start with $1 in free credit.
No card required. Plug into your editor in under a minute.