Modelpeers launches on Product Hunt today. It gives every AI model and agent a profile, the way LinkedIn does for people. You follow the models and agents you use, and their new versions, price changes, outages and retirement dates show up in one feed. If you build with AI, I’d love for you to support our launch and try the product:
Modelpeers records every public change to 200 AI models and agents from a dozen companies. Between October 15 and October 30 alone, it recorded 100 changes! Twenty were launches: new models, new versions and new features. Eighteen were major outages reported on official status pages. Thirteen changed prices, plans or usage limits. Eight were retirements and breaking API changes, and four switched a default model. The other 37 were smaller, like a model arriving on a new platform or a new benchmark result. On October 28 alone, GitHub Copilot changed 8 times.
Others see the same breakneck pace. BenchLM counted 314 AI model releases in the 12 months to October 28. That’s a new model about every 28 hours (!), and October was the busiest month on record, with 47 new models. ARK Invest found that the median gap between frontier releases from the five leading US labs fell from 37.5 days in 2023 to 11 days in 2026. What a time to be alive.
I’m not the only one who feels this way. Simon Willison co-created Django and writes one of the most-read blogs about AI models. In September, he told a developer conference: “As a software engineer, I’ve never had a year of my career where everything has changed so quickly and so dramatically.” He described one model as “clearly the best model in the world for a glorious eight days.” The same month, CNBC ran a story on “model fatigue”, quoting the CEO of Runpod: “I feel like model fatigue is a real thing.” On Reddit, someone asked if anyone else felt “a weird mix of ‘AI burnout’ and absolute awe.” Keeping up with “the daily firehose of new models, tools, and breakthroughs is starting to feel like a full-time job,” the post said.
Andrej Karpathy put it in one line last December, and 17 million people saw it:
Good news and bad news arrive the same way. A new model can make your product better overnight. A plan change can cut your usage in half. After a plan-limit cut on the morning of OpenAI’s DevDay in September, one AI commentator with 155,000 followers wrote: “$200 plan is now $100 plan, $20 is now $10. What the frick.”
I build with these models every day, and I still miss the important stuff. In April, a feature in TailoredRead, my site that writes nonfiction books with AI, was broken for two days before I learned that its model had been retired. Anthropic had announced the date two months earlier, on a page I never read, and I had no alert on that step. I wanted one place that tells me what changed in the tools I actually use. Modelpeers is that place.
I build with these models every day, and I still miss the important stuff.
What Modelpeers does
Modelpeers gives every AI model and agent a profile, the way LinkedIn gives one to a person. Today it covers 200 AI models and agents from 12 companies, including Claude, GPT, Gemini and Codex. You follow the ones you use, the way you follow people.
When one of them changes, the change shows up in your feed like a post. It can be a new version, a price or usage-limit change, a new default model, a major outage or a retirement date. Each one says what changed, who it affects and what to do, and links to the source.
Each profile shows what the model or agent is for and how its prices compare with its peers. It also lists its benchmark ranks, its incident history and short quotes from builders who use it. Each model was asked, through its own API, to describe itself and its weaknesses, and that answer is its “About me” section. The profile also shows the model’s writing habits, such as how often it uses dashes or “delve”...
The Upcoming list shows dates that are already announced, like retirements and price increases, with a countdown.
Once a week, on Monday, you get one email with only the changes for what you follow. If you want, urgent changes like a shutdown three days away reach you the same day. Comparison pages cover the pairs people search for most, like Codex vs Claude Code.
If you’d rather watch than read, here’s a one-minute video of Modelpeers.
A look under the hood
Every 4 hours, our backend reads 35 sources: makers’ changelogs, pricing and deprecation pages, official status pages, GitHub releases and public leaderboards. It drafts each new change with its quote. A script then fetches the source page again and validates every part of a change. A separate judge scores each change against its source, and a model from OpenAI checks any serious claim as a second judge.
Heads up: what’s about to change in AI
These dates are already on the Upcoming list. If you follow the models involved, each change reaches you before it happens.
November 12: OpenAI plans to end Cursor’s access to its models.
November 16: Copilot removes Claude Opus 4.8, Claude Opus 5 and GPT-5.6 Terra.
November 30: Anthropic retires Claude Sonnet 4.5.
January 1: Gemini 3.8 Flash and 3.6 Flash prices double, from $0.75 to $1.50 per million input tokens.
Simon Willison’s best model in the world lasted eight days. The models will keep changing every few days, and knowing what changed should not take a full-time job.
Modelpeers is live on Product Hunt today, and I’ll be in there all day answering every question and comment.
Which model or agent would you follow first, and what’s missing? Join the conversation: