Know the minute a new AI model drops.
A free Chrome push within about a minute of a new model going live on a provider's API, plus an email. Texts or a phone call if you want it anywhere.
What just dropped. What lands next.
Every launch we’ve called, newest first, then the ones we’re armed for, with live Polymarket odds where a market exists. Tap a model for its page.
One sweep every minute. Then everything fires.
No dashboards to babysit. Pick the models, pick the channels, and go back to work. The board does the watching.
Checkers sweep the APIs
Every 60 seconds we poll the provider APIs we can reach directly plus the public catalogs every lab ships to. A new model has to show up on two sweeps in a row, and where we can call it, it has to answer. Debounced, so no false alarms.
Every channel you enabled
The moment a launch is confirmed, your channels fire: Chrome push for everyone, SMS and a phone call on paid plans. Free email follows in a batch about 15 minutes later.
Then the intel rolls in
Every alert email carries a 30-second brief on the model. On paid plans, your own saved prompts run against it the day it lands, and we keep watching the subscriptions you pay for: the models they offer, their prices, their limits, and what quietly gets retired.
Everything on the watchlist.
The radar shows what lands next. This is the whole book: announced, leaked, rumored, or simply inevitable. Every one of them is swept every 60 seconds, and every one fires your alerts the second it answers.
Start free. Get louder when it matters.
Free is the Chrome push the second a launch is confirmed, plus an email about 15 minutes later. Pay only if you want your phone to buzz (Signal, $12, a text) or ring (The Call, $19.99, a call) that same second. No card to start, cancel anytime.
- Email alert on every launch
- Chrome desktop notification
- Dashboard: pick your models — SOTA-only, every lab, or a custom list
- The radar, the wire, the bench index + a page for every tracked model
- Live Polymarket odds on every watched launch
- Email arrives in the free batch — about 15 minutes after members' alerts
No card. No spam. Free forever. Instant push; email fifteen minutes later.
- Everything in Free
- Instant everything — email and SMS fire the moment a launch is confirmed, ahead of the free email batch
- SMS the minute a model goes live
- Per-model text routing — follow the models you care about, mute the rest
- Quiet hours — overnight texts hold until morning
Every feature on this card is live today. Cancel anytime. One tier up, your phone doesn't buzz — it rings.
- The automated phone call — a text gets missed, a call gets answered; it rings until you know
- Everything in Signal — instant SMS, per-model routing, quiet hours (inside your window the call rolls to a text, never a 3am ring)
- Ring, text, or both — chosen per model: ring for Gemini 4, text for the rest
- Charter rate — $19.99 never rises while you stay subscribed
Less than one of the AI subscriptions it guards. Every feature listed is live right now — nothing here is a promise. If a flagship drop doesn't ring your phone, that month is on us. Cancel anytime.
- Everything in The Call — the ring, the texts, the routing, the quiet hours
- Your prompts, scored on every launch — up to 20 of your real prompts run on the new model and on the model you use now; a judge from a third lab picks the better answer blind, the order alternating so neither side sits first
- The verdict first — “7 better, 2 same, 1 worse than yours” in the subject line, then the judge's reason, prompt by prompt
- Your report as a PDF — every prompt with the verdict, the judge's reason and both answers side by side (long ones abridged), plus reply times and the launch numbers
- Alerts built your way — pick the model to compare against, choose what the email carries (price, context, cost vs yours, vs today's #1 and the best value), and keep texts short or put the numbers in them
- Subscription watch — models quietly added to, or pulled from, the Gemini app & Anthropic's published model table (more surfaces next)
- Sunset & deprecation alerts — hear before a model you rely on is retired
- Price, limit & market-threshold alerts across the providers you pay
- The weekly intelligence brief — what changed and what to do about it
Everything above is live today. $34.99 is the charter rate and never rises while you stay subscribed. Cancel anytime. Already on Signal or The Call? Upgrade from your dashboard: your plan ends when the Watchtower starts, and its unused days are credited. Everyone already on a paid plan keeps every feature they have today, permanently.
✳ The Call is the alert you cannot miss; the Watchtower is the verdict on what landed. Every untagged feature above is live today — see the full roadmap → · our public track record →
The honest answers.
How fast is “the minute it drops”?
Our checkers sweep provider APIs every 60 seconds, and a detection is confirmed on a second sweep before anything fires. The moment it's confirmed, paid alerts fire instantly: Signal texts you; The Call rings your phone (and texts). Chrome push goes out to everyone at that same moment; free email follows in a batch about 15 minutes later.
Which models do you watch?
The radar pins the latest confirmed drops on top and shows what lands next — the next Claude Opus, Grok 4.8, GLM-5.5 and Kimi K3.1. Gemini 3.5 Pro sits there too, for a different reason: it has been in limited preview since July, and since Gemini 4 Argon was announced on Sep 30, 9to5Google reports it cancelled, though Google has said nothing. The full watchlist tracks 170, including the restricted Gemini Flash Cyber, plus Grok 5, Fable 6, Gemini 4, DeepSeek V5 and more. Click any model for its own page: the leaks, the news, the official word, and the live Polymarket odds on when it drops. Tune alerts to SOTA-only, every major-lab LLM, or specific models, plus extra streams for AI tools, local models, and Chinese models.
How do you avoid false alarms?
A detection has to show up on two sweeps in a row, in a provider’s own model list or a public API catalog. Not a rumor, a leak, or a docs page. Where we can call the model, it also has to reply to a test prompt before any channel fires; where we can’t, the alert email says so plainly, so you know what was checked and what wasn’t.
What is the subscription watch?
A Watchtower feature that watches the AI subscriptions you already pay for. Today that’s the published Claude model lineup and the Gemini app’s model list, plus API prices and limits across providers (ChatGPT’s surface is next). When a model is quietly added (or pulled, the way Fable 5 was in June), you hear immediately. Every paid plan (Signal and The Call) gets the subscription watch today, and keeps it.
What’s the difference between Signal and The Call?
Signal ($12) kills the delay: instant SMS on the same sweep that catches the launch, routed per model. The Call ($19.99) is the tier whose phone rings, because a text gets missed and a call gets answered. Both cost less than one of the subscriptions they guard. Both also get the Watchtower’s watchers: your prompts run on every new model day-0 (the raw answers), and alerts when your AI subscriptions add or pull models, change prices or limits, or retire a model you rely on. You keep them. What only the Watchtower adds is the scoring.
What’s the Watchtower tier?
The tier for people who want their own evidence. The Call tells you a model landed; the Watchtower tells you whether it’s better at your work. Up to 20 of your own prompts run on every new model and on the model you use now. A judge from a third lab (never the lab of either model) compares each pair blind, and the order alternates so neither answer always goes first. You get the verdict by email (how many better, same and worse, and why) plus a PDF with both answers side by side. You also choose what your alerts carry: the model to compare against, the price and context numbers, and whether your texts include them. It includes everything in The Call and the watchers. $34.99/mo is the charter price and never rises while you stay subscribed. Already on Signal or The Call? Upgrade from your dashboard: your current plan ends when the Watchtower starts and its unused days are credited, so you never pay for both. Anyone already on a paid plan keeps every feature they have today, permanently.
Can I cancel anytime?
Yes. Every paid plan is monthly with no lock-in. Cancel in one click and keep access through the period you’ve paid for. One caveat worth knowing: The Call’s $19.99 charter rate is locked while you stay subscribed. Cancel and the lock is gone; rejoining later means whatever the price is then. (Grandfathered founders: same deal on your $9.)
What happened to Founding?
Founding closed in July 2026. It was the launch-era tier, $9/mo for the early believers, and everyone who joined keeps that rate for life, grandfathered unconditionally. It closed as a decision, not a sell-out. The ladder today is Free, Signal, and The Call.
Do you have a desktop or mobile app?
On the way. The plan: a macOS menubar radar with a full-screen drop banner, then iPhone and Android apps with widgets, then Windows and Linux trays. Today, email works everywhere and Chrome push works in desktop browsers and on Android; on an iPhone, email (or a paid text or call) is how alerts reach you. See the roadmap.
Are you affiliated with OpenAI, Anthropic, Google…?
No. ArtificialWatch is independent and not affiliated with, endorsed by, or operated by any AI lab. We watch their public APIs and tell you the second something ships.
You won’t be watching when it drops. We will.
Claude Haiku 5.5 dropped Oct 7 at $0.10 / $0.50 per million for prompts up to 100K tokens, a tenth of Haiku 4.5’s price, with a 1M-token context window. Anthropic’s own table puts it ahead of GPT-6 Luna, OpenAI’s model at the same price, on every benchmark both were run on. Next on the radar: Gemini 4 Argon, announced by Google on Sep 30 for trusted testers, with no public API yet. Be on the board before it answers.