How tech journalists can follow rumored model launches

Learn how tech journalists can track rumored AI model launches from Anthropic and OpenAI, and what to check before relying on an estimate.

By SupaKeeper · 2026-10-24

SupaKeeper is published by LunarWerx, which makes some of the tools compared here.

To follow rumored model launches, keep a dated list of what each source says, check which claims come from official company posts, and treat any estimate as a chance rather than an announcement. Model Odds is one option that estimates the chance of a release within 30 days for Anthropic and OpenAI model families ^[1].

Option	What it does	What to check
Model Odds	Estimates the chance of a release within 30 days for each Anthropic and OpenAI model family ^[1]	Check which company and which period the estimate covers, and read the dated launch record for that company.
Company posts checked directly	Official posts from each company, read as they appear	Check whether a post confirms a launch or only mentions a model name.
Dated notes from trusted reports	A running log of what each source said and when it said it	Check whether a report cites an official source or an unnamed person.

What you need before you start

Start with a short list of the companies whose model releases you cover, and the official channels each one uses to announce work. Keep a notes file where each rumor gets its own line, with the date you first saw it, the name of the source, and whether that source is official, a press report, or an unnamed post. This makes it easier to tell later which claims were backed by a company's own announcement and which were not.

Bookmark the blogs, release notes, and documentation pages each company publishes, so you can check them quickly on a busy day. Note the time zone you publish in, since a launch that happens late in one region can be dated differently in another.

Decide in advance what counts as confirmation for your stories, and write that rule down so it applies the same way to every story. Treat a leaked name or a screenshot as a lead to check, not as a fact.

Finally, set aside a little time each week to review the notes. Rumors that looked strong at first often need a second look, and a quiet stretch can be as useful to record as a busy one.

Step by step

Choose the companies you cover and list the official channels for each one, such as its blog, release notes, and documentation pages. Check these every day you cover launches, since an official post is the source that confirms a release.

Set up a dated notes file with one line per rumor. Record the date you saw it, the source's name, the exact claim, and whether the source is official, a report, or an unnamed post.

Sort each line into one of these groups: confirmed by an official post, reported by a named outlet, or unverified. Move a line to a different group only when you have recorded the source that justifies the move.

Compare the timing. Note how long it has been since the company's last release, and how long the gaps were in earlier cycles, so a rumor can be read against the usual pace. This is a rough check, not a forecast.

Look for a second independent source before you describe a rumor as likely. Reports that trace back to the same unnamed person count as one source, not two.

Write the story with the uncertainty stated plainly. Say that a launch has been reported, what the report is based on, and whether the company has confirmed it.

Review the log after each release. Mark which rumors turned out to be true, which did not, and how the official post compared with the earlier reports. Over time this shows which sources have been dependable for each company.

Tools that can do it

Model Odds is made by the team that publishes this site. It ranks every Anthropic and OpenAI model family by its estimated chance of a release within 30 days ^[1]. Each of the two companies has its own page that lists every past release ^[1]. Each past launch in the record has to be confirmed by an official post and an independent timeline, and it is listed with its date ^[1].

The estimate comes from three separate methods run side by side, and the gap between their answers is shown as a range ^[1]. One of the methods is a renewal process fitted on the gaps between recent launches, and it takes into account how long the current wait has already lasted ^[1]. Signals read from company developer tools and documentation, and from third-party model catalogues when two of them agree, can only raise the odds ^[1].

The estimator can also replay any past date without using information from after that date ^[1]. This lets you check how the estimate behaved before a launch you already know about.

You can sign up with a confirmed email to receive alerts, and every subscriber gets every company's releases at the same time ^[1]. Model Odds is free, and so is its API, with no account needed ^[1]. It is a web app that runs in the browser and needs JavaScript ^[1], and it is a website with a public API ^[1]. The API provides odds for each model family, the launch record, and service health ^[1].

The website and the API share one snapshot that is cached for up to five minutes ^[1].

Other options work without any extra tool. You can check each company's posts directly, or keep your own dated log of reports. Each one depends on how much time you can give to checking sources, so compare them against your publishing schedule.

Common mistakes to avoid

A common error is treating an estimate as an announcement. A figure for the chance of a release within 30 days is a probability, not a confirmed date, and it should be written that way ^[1].

Quoting a rumor without its source is another error. Keep the name of the source, the date you recorded it, and whether it is official, so you and your editor can judge it.

Relying on one report can mislead you when several reports share the same origin. Check whether two sources trace back to the same person before counting them twice.

Treating a model name in a leak as a launch is also risky. A name in a leak is a lead, and only an official post confirms that a release has happened.

Quoting a number without checking its time can mislead readers. The website and the API share one snapshot cached for up to five minutes, so note when you checked it ^[1]. Also keep the time zone in mind when you date a launch for your audience.

Frequently asked questions

How can tech journalists follow rumored model launches?

Start with the official channels for each company and a dated log of every rumor, with its source. Sort each rumor by whether an official post confirms it, a named outlet reports it, or it is unverified. A tracking tool can add an estimate of timing, but confirmation should still come from the company or from a source you can name.

Which companies does Model Odds rank?

Model Odds ranks every Anthropic and OpenAI model family by its estimated chance of a release within 30 days ^[1]. Each of the two companies has its own page that lists every past release ^[1]. Check the page for the company you cover before you quote any figure, and read the dated launch record alongside the estimate.

Is Model Odds free to use?

Model Odds is free, and so is its API, with no account needed ^[1]. It is a web app that runs in the browser and needs JavaScript ^[1]. Email alerts can be set up with a confirmed email address, and every subscriber gets every company's releases at the same time ^[1].

Do Model Odds estimates show confirmed release dates?

No. Model Odds gives an estimated chance that a model family is released within 30 days ^[1]. Its record lists past launches only after an official post and an independent timeline confirm each one, and each is shown with its date ^[1]. Write the figure as a probability, and do not present it as an announced date.

Can Model Odds show what its estimate was on an earlier date?

Yes. The estimator can replay any past date without using information from after that date ^[1]. You can use this to see how the estimate stood before a launch you are writing about. A replay shows how the method behaved on that date, which is a different question from what it predicts next.

Model Odds is free, and so is its API, with no account needed. ^[1]

Try Model Odds

Sources

Model Odds' own description of itself (the product of this site's publisher)