AI aggregator
AI aggregator: every model on the leaderboard, one API
An AI aggregator, or LLM aggregator, gathers models from several providers behind one API. The Smart Router puts OpenAI, Anthropic, Google, DeepSeek, Moonshot AI and Z.ai behind one OpenAI-compatible endpoint, with your own provider keys and no markup on tokens, and goes one step further: it picks the model for each request from public benchmark results.
1,000 routed requests a month free · your own provider keys · no card
- 6
- providers behind one OpenAI-compatible endpoint.
- 25
- models, every one tested on the same benchmark tasks.
- 0%
- markup on tokens. Providers bill you directly, at their own list price.
01
What is an AI aggregator?
An AI aggregator, also called an LLM aggregator or AI API aggregator, puts models from several providers behind one API, one key and one log. Your code is written once against that API, and changing model or provider is a change of model name.
- One integration
- One SDK, one request format and one error format, whichever provider answers. The Smart Router speaks the OpenAI API, and Anthropic’s Messages API for Claude Code.
- One key in your code
- An aism_ router key reaches every provider you connected. The provider keys stay in the router, encrypted with AES-256-GCM, instead of in your application’s config.
- One log
- The model, provider, latency, tokens and cost of every request in one place, with exports and budget limits on the paid plans.
- Fallbacks across providers
- When a model fails, the next one answers, at another provider where you hold a key: up to 5 fallbacks on every request.
02
Credits or your own keys?
Aggregators bill in one of two ways, and the difference decides whose account the tokens run on.
- Credits
- You buy credits from the aggregator, it calls the providers on its own accounts, and usage is deducted from your balance: one bill, on the aggregator’s terms.
- Your own keys
- You connect your own provider accounts and the aggregator sends each request with your key. Each provider bills you directly at its own list price, under your own agreement, limits and data terms with it.
The Smart Router always works with your own keys, at every volume. Nothing is resold and nothing is added per token; plans count routed requests. Free includes 1,000 routed requests a month on every provider and every mode, with your own provider keys and no card, and the paid plans add volume, longer decision logs, API monitoring and team features.
03
Which models does the Smart Router aggregate?
Every model on the leaderboard, read from the roster as this page was rendered. Each one has its own page with its scores, its history and its drift detectors.
| Provider | Models on the leaderboard | API status |
|---|---|---|
| OpenAI | GPT-5.5, GPT-5.6 Luna, GPT-5.6 Sol, GPT-5.6 Terra, GPT-6 Astra, GPT-6 Luna, GPT-6 Sol, GPT-6.1 Sol, GPT-6.1 Terra | OpenAI status |
| Anthropic | Claude Fable 5.1, Claude Opus 4.6, Claude Opus 4.7, Claude Opus 4.8, Claude Opus 5, Claude Opus 5.5, Claude Sonnet 4.6, Claude Sonnet 5, Claude Sonnet 5.5 | Anthropic status |
| Gemini 3.5 Flash Lite, Gemini 3.8 Flash, Gemini 4 Argon | Google status | |
| DeepSeek | DeepSeek V4 Pro, DeepSeek V4.1 Flash | DeepSeek status |
| Moonshot AI | Kimi K3.1 Pro | Moonshot AI status |
| Z.ai | GLM-5.3 | Z.ai status |
04
An AI aggregator that also picks the model
Aggregation puts every model behind one API; the Smart Router also decides which one answers. Its 14 auto-* modes rank on the public boards, re-run on a fixed schedule: coding every four hours, reasoning and tool use daily, with an hourly canary probe.
auto-coding- The Coding board: repository bug fixes graded by running the project’s own tests.
auto-reasoning- The Reasoning board: long, multi-turn working sessions.
auto-tooling- The Tool use board: sandboxed agent tasks with real tools.
auto-value- Among the models within 5 points of the best Combined score, the most points per dollar of a scored run.
auto-task- Decided per request from what it contains: code, tools, design and planning, or long text.
The methodology is public: every weight, statistical test and drift constant. The test tasks are private, so no lab can train on them. The router’s live pick for each mode is on AI router, and every mode with its rule on the Smart Router page.
05
Start in three steps
Add your provider keys, create a router key, change the base URL.
- 1. Add your provider keys
- Create a free account and add a key for each provider you use. Each one widens the set of models the modes can rank.
- 2. Create a router key
- One aism_ key for every provider and every mode, used wherever your code expects an OpenAI key.
- 3. Change the base URL
- Point any OpenAI-compatible client at
https://aistupidlevel.info/v1, and send a mode or a model id as the model.
curl https://aistupidlevel.info/v1/chat/completions \ -H "Authorization: Bearer aism_your_key_here" \ -H "Content-Type: application/json" \ -d '{ "model": "auto-best", "messages": [{"role": "user", "content": "Summarise this contract in five points"}] }'
Guides for Claude Code, Cursor, LangChain, n8n and more are on the integrations hub; every endpoint is in the router documentation.
06
AI aggregator questions
What is an AI aggregator?
An AI aggregator is a service that puts models from several AI providers behind one API, one key and one log, so an application is written once and can call any of them. The AI Stupid Level Smart Router aggregates OpenAI, Anthropic, Google, DeepSeek, Moonshot AI and Z.ai behind one OpenAI-compatible endpoint, with your own provider keys and no markup on tokens, and also chooses the model for each request from public benchmark results.
What is an LLM aggregator or an AI API aggregator?
The same thing under other names. An LLM aggregator puts large language models from several providers behind one API; an AI API aggregator stresses that the providers’ APIs are what is being combined. Either way, your code speaks one request format and the aggregator speaks each provider’s.
Is an AI aggregator the same as an AI router?
An aggregator gives you access: many models behind one API. A router adds the decision: which of those models answers each request. The Smart Router is both. Its auto-* modes rank on the public boards, so a request sent as auto-coding goes to the model placing highest on the Coding board among the providers you hold keys for. More on what an AI router is.
Do I need an account with every provider?
You add a key for each provider you want in play, and the modes rank only the models those keys reach. Two keys give you a router across those two providers’ models, and more keys widen it. Each provider bills you directly, under your own terms and limits with it, and the keys are stored encrypted with AES-256-GCM.
Does an AI aggregator cost more than calling providers directly?
Not on tokens with the Smart Router: requests go out with your own keys and each provider bills you at its own list price. Plans count routed requests instead. Free includes 1,000 routed requests a month on every provider and every mode, with your own provider keys and no card. Pro Intelligence $12/mo for 10,000 routed requests a month; Developer $25/mo for 100,000. Past an allowance, top-up credits at $1 per 2,000 requests, minimum $5, that do not expire. Plans and prices.
Which AI aggregators and routers are there?
Besides the Smart Router, the AI routers and gateways in the same space include OpenRouter, Requesty, Vercel AI Gateway, LiteLLM, Portkey, Cloudflare AI Gateway, Helicone AI Gateway and Not Diamond. Best AI routers in 2026 compares them from each vendor’s own pages, and OpenRouter alternatives adds the steps to switch.
Every model on the leaderboard, one key
Free includes 1,000 routed requests a month on every provider and every mode, with your own provider keys and no card.
1,000 routed requests a month free · your own provider keys · no card
Coding, Reasoning, Tool use and Combined, over 24 hours, 7 days and 30 days.
Already have an account? Log in