Decision models

Decisions in one forward pass

Models that read a state and a schema of typed questions and return a probability for every option. No text generation, no parsing.

How it works

  1. Send a state. A document, an email, a JSON record, up to 8 images or the frames of a short video.
  2. Pick the questions. A ready use case fixes typed questions (choice, score or yes/no) with their options, or you send your own.
  3. Get probabilities. One forward pass returns a probability for every option of every question, plus a confidence.
  4. Act on the band. act runs automatically, confirm asks a person, human hands over.

Use cases, ready to call

Core decisions

The nine building blocks: routing, guardrails, tool gating, triage, reranking, evals, labeling, real-time control and confidence gates.

Accounts payable

For finance teams and bookkeeping bureaus: sort incoming documents, catch duplicate invoices, decide approve / hold / escalate and code costs with VAT treatment.

Agent safety

For teams that run AI coding and ops agents: decide whether a tool call, shell command or code change may proceed.

Clinical intake

Synthetic patient messages (EN, SL) triaged for urgency, routing, red flags and clinician review. Decision support only, not a medical device: a clinician owns every clinical decision.

Cold-chain monitoring

For pharmacies, hospitals, labs and food logistics: triage temperature excursions, spot faulty sensors, decide what happens to the stock and who gets woken up.

Outreach & content

For sales, marketing and community teams: classify outreach replies, grade posts, flag low-quality AI text and score lead fit.

Content moderation

Trust and safety teams decide what stays up: action on posts, policy labels, appeal gate and marketplace listing checks.

Contract review

Customer-side review of supplier contract clauses: label each clause, score its risk against a playbook, find the clause that answers a question and decide accept, negotiate or reject.

Osnovne odločitve (SL)

Slovenska zrcalna različica paketa core: istih devet gradnikov z enakimi primeri in oznakami, za neposredno primerjavo istega modela v angleščini in slovenščini.

Data flywheel

Curate and judge LLM training data: triage rows, pick DPO preferences, grade answers against sources and label rows for the mix.

Dev & ops triage

For SRE and platform teams: triage log lines, vet dependencies, classify CI failures and route issues.

Education and tutoring

Tutors and teachers of school-age learners grade answers, spot misconceptions, decide when to give a hint and read the mood of a message.

Games & real time

For game studios and live-service teams: tiny decisions that must be answered in milliseconds: cheat suspicion, player intent, difficulty and board moves.

Industrial faults

For maintenance and plant teams: diagnose machine faults from telemetry, prioritise maintenance, validate sensors and sort alarm floods.

Market intelligence

Analysts and trading desks triage crypto, equity and commodity news: catalyst strength, source status, tags and thesis-relevant headlines.

Public inbox

Citizen e-mails to a municipality (SL, EN): route to the right department, recognise the request type, set the response deadline class and flag personal data.

Research & documents

For analysts, knowledge teams and data teams: fact-check claims against evidence, grade answers with a rubric, tag notes and route questions safely to a data layer.

Security operations

For SOC analysts and MSSPs: give alerts a verdict and severity, gate response actions, triage reported e-mails and label log lines.

Trading compliance

Market-conduct surveillance for energy and financial trading desks: inside-information checks, suspicious order patterns, trader chat review and regulatory report routing.

Images & video

Decisions from a photo, a screenshot, a chart or a short clip (frames): document intake, photo quality, screenshot triage, chart reading, motion and status lights.

Working time & absence

For HR, payroll and shift planners: classify absence messages, check timesheets against EU-style rest, break, night and overtime rules, and route corrections and shift swaps.

Plans

One key, one meter: input tokens (about 400 for a typical decision, so 100 M tokens are roughly 250,000 decisions). No output tokens. Prices in euro, without VAT; Slovenian VAT of 22 % is added at checkout.

Free
€0
  • The live demo, no sign-in
  • All packs, results stay in your browser
  • Per-visitor limits
Developer
€19 / month
  • 100 M input tokens a month
  • All packs on the fast model
  • 120 requests a minute, 2 at once
Team
€99 / month
  • 1,000 M input tokens a month
  • Runs ahead of Developer in the queue
  • 600 requests a minute, 4 at once
  • One custom pack a year, calibrated on your examples (on request)
  • Email support within 2 working days
Business & private
Talk to us
  • EU-only processing, data processing agreement, nothing stored
  • A dedicated deployment by arrangement
  • Custom packs, quarterly accuracy report

Plans and keys live on your MediaAtlas account, and the same key works for every MediaAtlas service. Past the monthly limit the key goes on paying from your credit at €0.28 per million input tokens (no plan needed for that); switch it off on the key and the API answers 429 quota_exceeded until the 1st instead. Got a discount code? Enter it on the payment page. You pay Stripe; card details never reach us. Cancel or change plan any time in the billing portal.

For agents and bots

An agent calls the API with your key, like any program. When the plan needs paying, the API answers 402 payment_required and an agent can ask for a payment link (POST /api/checkout) to hand to a person. Download the agent skill or read the API documentation.

Questions

How is this different from prompting a chat model?
A chat model writes text that you then parse, and its stated confidence is a guess. A decision model scores every allowed option directly, so the answer is always one of your options and the probabilities can drive thresholds, routing and audits.
What can it read?
Text and JSON up to 20,000 characters per input, up to 32 inputs per call, and images or video frames for use cases that read them.
Can I ask my own questions?
Yes. Send a questions schema instead of a pack and use case; each question has a type, an instruction and criteria per option. See custom questions.
How is it billed?
By input tokens, in monthly plans (see plans). There are no output tokens: the model does not generate text. Or without a plan, from prepaid credit at €0.28 per million. Your keys are on your MediaAtlas account.