Track AI models, prices and retirements

Models launch, change price and retire every week. Know the day a cheaper model appears, a price moves, or a model you depend on gets a retirement date.

For aI teams, platform engineers, finance owners of AI spend, and agents that choose models.

Get these events How it works

The events

What your agent listens to

  • AI model released

    ai_model.released

    Ready Public AI model catalogue, worldwide

    New model: Atlantico Vision 7B ($0.10 in, $0.40 out per million tokens)

  • AI model price changed

    ai_model.price_changed

    Ready Public AI model catalogue, worldwide

    Price cut: Atlantico Vision 7B output now $0.30 per million tokens (-25%)

  • AI model retirement announced

    ai_model.retirement_announced

    Ready Public AI model catalogue, worldwide

    Atlantico Vision 6B retires on 2026-12-31

  • AI model retired

    ai_model.retired

    Ready Public AI model catalogue, worldwide

    Retired: Atlantico Vision 5B is no longer offered

How it works

From event to action

  1. A model appears, changes price or gets a retirement date.
  2. Your agent compares it with the models your product uses.
  3. It estimates the cost change on your traffic and proposes a switch.
  4. Retirements become tickets with the deadline before anything breaks.

Wake your agents up. Each event can start Claude Code, Codex or your own script the moment it lands, one by one or in batches: no polling, no cron, no dashboard to check.

Events carry text from public sources. Run agents that act on them isolated, with only the access the task needs. Run them safely

Create your own stream

Tell your agent
npx worldsignals watch --type ai_model.released,ai_model.price_changed,ai_model.retirement_announced,ai_model.retired \
    --agent claude --prompt ./brief.md

# brief.md
When a model we use gets a retirement date or a price change over 10%, open a ticket with the date, the cost impact and the closest replacement.

Questions

About this use case

Where does the model data come from?

A public catalogue of hundreds of models across providers: names, prices per million tokens, context sizes, modalities, release and retirement dates.

How often is it checked?

Every hour. Each price change is its own event with the previous price.

Do you include open-weight models?

Yes, and events say whether the weights are published.

More use cases

Give your agents something to notice

Join the waiting list, free during early access. Pick these events and we’ll email you when you’re in.

Get these events See the live stream