unbiased ai

Dirac, running on Pareto

Dirac is built for exact changes in real repositories: hash-anchored edits, syntax-tree inspection, subagents and a plain mode you can script. It also lists Unbiased as a provider. Give it your Unbiased API key and the model id pareto, and each request goes to Pareto, which selects frontier and open models for the job and blends their outputs into one answer.

Choose Personal · $10/month See the setup →

Tested live with Dirac CLI 0.5.2250M tokens a week

A demonstration of Dirac on Pareto, recreated from a real run. Not a screenshot.

Set it up with one command

Dirac has Unbiased built in as a provider (id unbiased), already pointed at the Unbiased API. Save your key once with dirac auth, or pass it as DIRAC_* environment variables for a single run.

  1. Install Dirac

    Run npm install -g dirac-cli. Dirac needs Node.js 22.13 through 24.x; its README says Node 25 is not supported.

  2. Get an Unbiased API key

    Choose a plan, then create a key in the Unbiased platform. New accounts are reviewed and approved by a person before the key works.

  3. Choose the Unbiased provider

    Run dirac auth --provider unbiased with your key and the model id pareto. There is no base URL to enter: the provider already points at https://api.unbiased.ai/v1.

  4. Give it a task

    Run dirac for the terminal UI, or dirac --yolo "…" for plain output with every action auto-approved.

Tested on October 20, 2026Dirac CLI 0.5.22 on Pareto (pareto). It found the off-by-one in a failing Python test, edited the file and reran the test to green. Run record.
Save it once
npm install -g dirac-cli
dirac auth --provider unbiased \
  --apikey your-unbiased-api-key \
  --modelid pareto
dirac
Or for one run
DIRAC_PROVIDER=unbiased \
DIRAC_MODEL=pareto \
DIRAC_API_KEY=your-unbiased-api-key \
dirac --yolo "Run the tests and fix failures"

The Unbiased provider arrived in Dirac 0.5.17. On an older version, or if you prefer it, use the OpenAI Compatible provider instead: --provider openai --baseurl https://api.unbiased.ai/v1 --modelid pareto, including the /v1. Type pareto by hand there, since the Unbiased API does not serve a model list for Dirac to pick from. In the VS Code extension, open the Dirac sidebar and choose Unbiased as the API provider, or OpenAI Compatible with the same base URL, key and model id. Dirac saves provider settings, key included, in plain text under ~/.dirac/data/. Set DIRAC_DIR to keep them somewhere else.

Dirac stays. The model changes.

Dirac keeps doing the careful part. Pareto does the thinking behind it.

One model string, several models behind it

Pareto is Unbiased’s blended AI model. On each request it selects frontier and open models and blends their outputs into one answer, so Dirac talks to one model string, pareto, and never has to pick.

“When the frontier improves, Pareto improves; when frontier prices fall, you should benefit.” From the pricing page.

How a Pareto answer gets made →
An illustration, not a wiring diagram. The blend includes frontier and open models, and its composition can change.

What an agent task costs on Pareto

Agents make many model calls per task, so the price of a finished task matters more than the price of a token. The model card publishes both for Pareto on two public agent benchmarks.

Terminal-Bench 4.0Tasks an agent completes in a terminal
$0.48/task
50.8% solved
DeepSWE v1.1Changes an agent makes in a real codebase
$0.24/task
69.9% solved

Pareto 26.10, preliminary October 1, 2026 runs that may change; denominators and the cost method still require confirmation. Cost is the reported mean USD per task at API rates, not an estimate for your workload. Model card and method.

Pay for Dirac by the month

One API key works across Dirac and any other tool you point at Pareto. Pick a monthly allowance, or pay per token on Team.

Personal

$10/month

  • 50M tokens a week
  • 20 requests per minute
  • For one person
Choose Personal

Personal Max

$100/month

  • 600M tokens a week
  • 60 requests per minute
  • For one person
Choose Personal Max

Team

Pay as you go

  • Prepaid credits at published rates
  • 300 requests per minute
  • No weekly token cap
Get started with Team

Subscriptions are for one person and are not unlimited. New API accounts are reviewed by a person before access turns on, and commercial or production use needs a separate written agreement. Rates and limits are on Pricing.

Dirac on Pareto, questions answered

Which Dirac provider should I pick?

Unbiased (id unbiased), on Dirac 0.5.17 or later. We ran Dirac CLI 0.5.22 against Pareto on October 20, 2026 with both the Unbiased provider and OpenAI Compatible (id openai, base URL https://api.unbiased.ai/v1): each time it found the off-by-one behind a failing Python test, edited the file and reran the test to green. The Unbiased provider needs no base URL. Format compatibility is not a promise that every Dirac feature works, so tell us if one doesn’t.

Why does Dirac say the context is 262K?

Dirac keeps its own record of model limits and listed Pareto at 262,144 tokens of context and 131,072 of output, so its meter reads against 262K. Its cost readout also uses Dirac’s own price table. Your Unbiased plan, not that figure, decides what you pay.

Does all of Dirac’s traffic go to Pareto?

The main agent’s does. Dirac also has an optional Utility model for jobs like compaction and permission decisions. If you set that to another provider, those requests go there instead.

Is it safe to run --yolo?

--yolo approves every action without asking, which is what makes it useful in scripts. Use it in repositories you trust, since a project can bring its own hooks and tools. Without it, a standalone task stops with an error when it needs your approval.

Can I switch back?

Yes. Run dirac auth again and choose another provider, or override a single run with --provider and --model.

Is Pareto made by the Dirac team?

No. Pareto is Unbiased’s model. Dirac is made by Dirac Delta Labs. Unbiased sponsors Dirac’s open-source development, and Dirac lists Unbiased as a sponsor in its README. You keep Dirac and point it at the Unbiased API.

What happens when I hit the weekly allowance?

Requests over your plan’s limits return an HTTP 429 error. Personal includes 50M tokens a week and 20 requests per minute. Personal Max raises that to 600M a week and 60 requests per minute for $100/month. Team uses prepaid pay-as-you-go credits with no weekly token cap.

Do I choose between models?

No. You send every request to one model string, pareto. Pareto selects frontier and open models for each request and blends their outputs into one answer.

Is my code kept?

Personal and Personal Max are not our zero data retention tier: prompts and responses may be kept up to 30 days, as section 5(c) of the Terms explains. For work that must not be kept, ask about the Zero Data Retention API on the Private AI page.

Pareto works in other tools too

The same key and the same model string go into each of these. See them all.

Product names belong to their owners. Except where a guide says otherwise, Unbiased is not affiliated with or endorsed by the makers of the tools on this page.

Hand Dirac a bug with a failing test

One dirac auth line, one key and the same exact edits. Pareto starts at $10 a month for 50M tokens a week.

Choose Personal · $10/month See the setup →