Sweep
First-party /v1/models endpoints every 10 s, OpenRouter every 20 s, docs pages every 30 s, Hugging Face and lab blogs every minute. Unchanged catalogs cost a 304.
Connecting to the watcher…
Know the second a new AI model drops.
Instant Model sweeps lab APIs, OpenRouter, Hugging Face and docs pages every few seconds. The moment a new model shows up, your phone buzzes — or rings.
- —sources live
- —fastest sweep
- —model IDs tracked
- —last sweep
Live drops
Everything the watcher has spotted, newest first. New drops appear here the instant they're detected.
When will it drop?
Live release odds from prediction markets, leaks and "coming soon" reports, and code that lands before a launch.
Latest news
Early signals
Model support often lands in vLLM, transformers, llama.cpp or SGLang before launch. We flag it when the model hasn't shipped yet. Odds swings of 15+ points in a day show up here too.
Why it's instant
Labs flip their model lists before the tweet, the blog post or the livestream ends. We read those lists directly.
Diff & dedupe
claude-opus-4-5-20251101, anthropic/claude-opus-4.5 and the docs page are one model. You're alerted once, by whichever source saw it first. Snapshots and quantized re-uploads don't count as launches.
Ring
Every channel fires in parallel the moment a drop is confirmed: phone push, Pushover alarm, SMS, a real phone call. Delivery usually lands in under a second.
Set up your alerts
Three steps, about a minute. No account needed; your settings are tied to this browser.
Works with iOS Shortcuts
Shortcuts can't listen in the background, so the instant alert comes from a push app. Shortcuts is great on top of it.
"Hey Siri, any new models?"
- New shortcut → Get Contents of URL:
/api/latest.txt - Add Speak Text (or Show Result).
- Name it "Any new models". Siri reads you the latest drops.
Add ?companies=openai,anthropic to filter.
"Hey Siri, when will GPT-6 drop?"
- Get Contents of URL:
/api/odds.txt?q=gpt-6 - Add Speak Text. Siri reads the live release odds.
- For rumors instead:
/api/news.txt?topic=upcoming
Run a shortcut when a model drops
- Install Pushcut and create a notification; copy its webhook URL.
- Paste it into the Webhook & iOS Shortcuts channel above.
- In Pushcut, attach any shortcut to that notification: flash the lights, open the model page, start a timer…
Make it impossible to miss
- ntfy: alerts arrive as Time Sensitive, so allow ntfy in your Focus settings.
- Pushover: emergency priority repeats every 30 s until you acknowledge it.
- Phone call: add the caller number to Favorites so it rings through Do Not Disturb.
Plans
Every plan watches every lab. You pay for louder ways to be reached.
Watch
$0forever
- Phone push via ntfy, browser & home-screen push
- Telegram and Discord alerts
- All 20 labs, flagships or everything
- 10-second API sweeps
- Model news, rumors and live release odds
- Daily briefing at your time
Siren
$4/ month
- Everything in Watch
- Pushover emergency alarms that repeat until you acknowledge
- Hand-picked models and keyword filters
- Email alerts
- "Model pulled" alerts
Most popular
Wake-Up Call
$9/ month
- Everything in Siren
- A real phone call that reads you the model name
- SMS backup
- Rings through Do Not Disturb when saved as a favorite
Lab
$29/ month
- Everything in Wake-Up Call
- Webhooks: Slack, Pushcut, Home Assistant, Zapier
- Up to 10 channels per team
- JSON API and live event stream
Running your own copy? Every plan is unlocked and billing isn't switched on. SMS, calls and email use your own Twilio/Resend keys and need the admin code.
What we're watching
Live health of every source. Sources marked "needs key" switch on when the server owner adds a free API key.
| Source | Covers | Every | Last sweep | Status |
|---|
FAQ
How fast is "instant", really?
Detection time is the sweep interval plus a fraction of a second: with an API key a new model is caught within 10 seconds of appearing (about 5 on average), and the push reaches your phone in well under a second after that. Nothing can alert you before a lab publishes something, but provider model lists usually flip before the announcement goes out.
How do you know when a model will drop?
Nobody knows exactly, so we show the best public evidence: release odds from Polymarket (real money) and Manifold (play money); leaks and "coming soon" reports from the tech press, Hacker News and lab blogs; and early signals such as support for an unreleased model appearing in open-source inference engines. Turn on the daily briefing to get all three each morning.
Why not just an iOS Shortcut?
Shortcuts only run when something triggers them on your phone and can't poll every few seconds in the background. The watcher runs on a server and pushes to an app built for instant, Focus-breaking notifications (ntfy or Pushover). Shortcuts are still useful on top: see Works with iOS Shortcuts.
Will it wake me up?
Pick Pushover with emergency alarms (repeats until you acknowledge, and can use iOS Critical Alerts) or a phone call. ntfy alerts arrive as Time Sensitive, which gets through Focus modes you allow it in.
What counts as a new model?
In Flagships only, a new model family, like gpt-6 or claude-opus-5-5. Dated snapshots (-2026-09-01), preview tags, quantized re-uploads (GGUF, FP8, AWQ), embeddings, TTS and moderation models are skipped. In Everything, any model ID a lab hasn't listed before.
Won't I get spammed when the server starts?
No. The first sweep of every source only records a baseline. If a source suddenly lists dozens of unseen IDs (a catalog reset or format change), that's logged as a notice, not pushed as 40 alerts.
Does it need API keys?
No. OpenRouter, Hugging Face, docs pages and lab blogs work without keys. Free API keys for OpenAI, Anthropic, Google, Mistral, xAI and others add the fastest signal of all: the lab's own model list.