How to use More AI for free: OpenRouter’s free models and other zero-cost routes

You do not need a paid API account to start. Options include built-in starter access for eligible signed-in beta accounts, provider free tiers, OpenRouter’s :free models and local models. Here is what each route requires and where its limits are.

The More AI team · Updated September 27, 2026

What “free” actually means here

Free access can come from a provider that does not charge for a model, a model running on your own computer, or built-in starter models that More AI makes available to eligible signed-in users during the beta. Built-in availability and quotas follow the current account entitlement and may change with the rollout.

The routes are not equivalent. With sync, online tools, master telemetry and AI observability (including replay) off, a local model keeps prompts and inference on your machine but is limited by your hardware. Provider free tiers and built-in starter models are hosted and can have daily or account-level limits.

On a phone, you can open the web app and check which cloud features your account can access. The key, local-model and folder setup below describes the desktop app; running a local model on your computer still requires that computer.

  • OpenRouter — selected free API models, with account and model limits
  • Google AI Studio — free API access for eligible models, accounts and regions
  • Groq — a limited free API tier; check current model and account quotas
  • Ollama / LM Studio — local inference with hardware and setup requirements

An OpenRouter key and free model variants

Free-model availability and request/token limits change by model, account and provider. Read the official limits and the limits shown for your account before a long agent run. A free chat interface is not the same as a free API allowance.

  1. Create a free account at openrouter.ai and open the Keys page.
  2. Press “Create Key”, give it a name, and copy the value — it starts with sk-or-.
  3. In More AI open Settings → Providers (or the “Start free” group during first-run setup) and pick OpenRouter.
  4. Paste the key. The app immediately fetches OpenRouter’s live model list.
  5. In the model picker, search for “:free” to see only the models that cost nothing.
The model picker in More AI, searching the live list from a connected provider

Google AI Studio and Groq

Google AI Studio and Groq offer model-dependent free API access where available. Neither is universally the fastest or most generous: rate limits, region, context and workload determine what works. Check the current model and account limits, then try a small task.

Local models: setup and offline requirements

Start a local server before connecting More AI. Ollama uses http://localhost:11434/v1 by default. In LM Studio, load the model and start the local server, normally http://localhost:1234/v1; enter its token if you enabled server authentication. These local endpoints differ from the providers’ hosted services.

Local inference has no provider per-token bill. Hardware, electricity, storage and your time still have a cost. Capability depends on the exact model, quantization, context length and available memory; test one small task before choosing it for a workflow.

Where free runs out — and what to do then

Free APIs can have per-minute, token, daily or account limits. A document task can make several model and tool calls; its cost depends on context, output, retries and the provider. Check current limits before starting and record failures instead of treating an incomplete result as finished.

The check below is a repeatable procedure for your own run, not a published More AI model benchmark. It separates a chat response from the model actually reading a file through a tool.

  1. In More AI desktop, choose a model currently available on your plan. Create a separate folder with a file named more-ai-test.txt containing two lines: “T01, done, 2 hours” and “T02, pending, unknown hours”.
  2. Open that folder in Cowork and ask: “Read more-ai-test.txt. How many tasks are done, and is T02 actual time known? Do not change the file.”
  3. Check that the app actually read the file and answered: one task is done, and T02 actual time is unknown. If the model cannot call the needed tool or gets this wrong, do not move a real document workflow to it.
  4. Record the model, current account limit, any error and the request count in your provider dashboard. For a longer task, test cost and output separately using copies of non-sensitive files.

Frequently asked questions

Do I need an account with More AI to use free models?

Not for provider free tiers or local models in the desktop app. Built-in starter models require sign-in and the relevant account entitlement. An account is also required for cloud execution, synchronization, the web app and access from other devices.

Which free option should I start with?

If your signed-in account offers a starter model, that is the quickest no-key route. For the widest provider choice, use OpenRouter; for long documents, try Google AI Studio. For local prompts and inference, use Ollama with sync, online tools, master telemetry and AI observability (including replay) off.

How many free requests do I get on OpenRouter?

Check OpenRouter’s current limits page and your account. Free variants have request and daily caps; credit history and rollout can affect the allowance. An agent task may consume several requests.

Is a free model good enough for real work?

For drafting, summarising, questions and light editing, yes. For long multi-file code changes and hard reasoning, a paid frontier model is still noticeably better. You can switch per conversation, so use the cheap one by default and the expensive one when it matters.

Do free models see my files?

A hosted model sees whatever you send it, the same as any API. Provider terms on training use differ by tier. With Ollama or LM Studio, prompts and inference stay local when sync, online tools, master telemetry and AI observability (including replay) are off.

All guides