SupaKeeper Plain comparisons of AI coding tools

How to monitor model version changes across providers for benchmark research teams in 2026

A tool-neutral guide for benchmark research teams to track model version changes across providers, with options compared.

By SupaKeeper ยท

SupaKeeper is published by LunarWerx, the makers of AgentHydra, Copilot AI Productivity Suite, Codex Forecast and Model Odds, which are compared here alongside other tools.

To monitor model version changes across providers, gather the provider sources your team already trusts, record each change with its date, and check that record on a fixed routine.

OptionWhat it can doWhat to check
Model OddsRanks every Anthropic and OpenAI model family by its estimated chance of a release within 30 days. [1]Check that the providers and model families you track are among those it covers.
Provider announcements checked by handUses each provider's own posts as the record of what changed.Check how often the team reads them and whether each change is logged with a date.
Team tracking spreadsheetHolds a dated list of the versions your team has seen and benchmarked.Check who updates it, how it is shared, and whether it is linked to benchmark results.

What you need before you start

Before you start, decide which providers and model families your team needs to watch. Benchmark work usually depends on specific version names, so write down the provider, the model family, and the exact versions your past results were produced with.

Keep a dated record of the versions you have already benchmarked. Without a baseline, a change in a provider's model is hard to connect to a change in your numbers.

Have a list of the sources you trust for change notices, such as provider announcements, changelogs, and documentation pages. Know who on the team reads each source and how often.

Check that everyone who needs the record can reach the sources and the shared file where changes are logged.

Step by step

  1. Name the providers and model families your benchmarks depend on. Write each one down with the version names your past results used.
  1. Save a dated list of the versions you have benchmarked. Later changes need this baseline to compare against.
  1. Pick the sources you will watch for changes. Provider announcements, changelogs, and documentation pages are common choices. Check whether each source gives a date for every change it reports.
  1. Set a fixed check routine. Decide how often someone reads each source and who records what they find.
  1. Log each change when you see it. Note the provider, the model family, the version name, the date you saw the change, and the source it came from.
  1. Compare each logged change with your benchmark record. Decide whether any benchmark needs to be rerun before results are compared.
  1. Review the list of sources every few months. Providers and their version names can change, so the list needs regular upkeep.

Tools that can do it

Model Odds is made by the team that publishes this site.

Model Odds ranks every Anthropic and OpenAI model family by its estimated chance of a release within 30 days. [1]

It keeps a page for each of those two companies that lists every past release. [1]

Each past launch it uses has to be confirmed by an official post and an independent timeline, and it is listed with its date. [1]

The estimate runs three estimators side by side and shows the spread between them as the error bar. [1]

Model Odds bases its estimate on a renewal process fitted on the gaps between recent launches, and that estimate takes into account how long the current wait has already lasted. [1]

Signals read from company SDKs and docs, and from third-party model catalogues when two of them agree, can only raise the odds. [1]

The estimator can replay any past date without using data from after that date, which lets you review how it reached its estimates. [1]

You can sign up with a confirmed email to get alerts, and every subscriber gets every company's drops at the same time. [1]

Model Odds is a web app that runs in the browser and needs JavaScript. [1]

Model Odds is a website with a public API. An API serves per-line odds, the launch record and service health. [1]

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

Other approaches include keeping a record by hand from provider announcements, or maintaining a tracking spreadsheet your team shares. Compare each option against the job of monitoring model version changes across providers, and check what each one covers before you rely on it.

Common mistakes to avoid

Avoid treating an estimated chance as a confirmed date. Model Odds reports an estimated chance of a release within 30 days. [1]

Avoid assuming a figure is live. Model Odds shares one snapshot between its website and its API, and that snapshot is cached for up to five minutes. [1]

Avoid logging changes without dates or version names. A record that cannot be lined up with benchmark runs does not help when results are compared.

Avoid relying on one source. A single provider notice may miss a change that another source reports, so keep the list of sources current.

Frequently asked questions

How can benchmark research teams monitor model version changes across providers?

Benchmark research teams can monitor model version changes by naming the providers and model families they depend on, keeping a dated record of each version they have benchmarked, and checking provider announcements on a fixed routine. A single shared record reduces the chance that a change goes unnoticed. Check whether each source gives a date for every change, because undated notices make it hard to line up changes with benchmark results.

Does Model Odds cover all model providers?

Model Odds ranks every Anthropic and OpenAI model family by its estimated chance of a release within 30 days. [1] Check whether the other providers your team tracks are among those it covers before relying on it for them. A provider that is not covered needs its own source of change notices, such as its announcements or changelog.

Is Model Odds free for research teams to use?

Model Odds is free, and so is its API, with no account needed. [1] The website is a web app that runs in the browser and needs JavaScript. [1] Teams that want to pull the data into their own tools can use the public API, which serves per-line odds, the launch record and service health. [1]

Can Model Odds tell the exact date of a model release?

Model Odds gives an estimated chance that a model family releases within 30 days, which is an estimate rather than a confirmed date. [1] Its estimate is a renewal process fitted on the gaps between recent launches. [1] The spread between its three estimators is shown as the error bar, so the uncertainty stays visible. [1]

How can a team review what Model Odds based its estimates on?

The estimator can replay any past date without using data from after that date. [1] Each past launch it uses has to be confirmed by an official post and an independent timeline, and it is listed with its date. [1] Signals read from company SDKs and docs can only raise the odds. [1]

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

Try Model Odds

Sources

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

Related comparisons