Home
/
A single chronological feed, newest first. Each card links out to the official blog or paper when we have one, and carries a short description, modality, open- vs closed-weight signal, and known variants.
Manifesto
Why/this?
AI is accelerating at breakneck speed. Every week, a new wave of models drops from labs around the world — and keeping track of it all is nearly impossible. This site is one long answer to a simple question: what happened, when, and who shipped it?
2017Chronological coverage2026
The idea
This timeline covers major LLM and related model releases from the Transformer era to today — so you can visualize how fast we're moving. Each row ties a model to its lab, release window, modality, licensing posture, and (when available) variants and scale.
The companion graph turns the same dataset into swimlanes, growth curves, density maps, and more — for when you want patterns, not a scroll.
Built to earn the bookmark: minimal by design, fast by default, and always worth returning to.
What's inside
Timeline
Snapshot
/snapshot
The timeline data, crunched into numbers — how many models shipped, who's releasing the most, which months get crowded, open vs closed splits, and a bunch of calendar oddities you wouldn't notice just scrolling.
Graph
/graph
The same records, reorganized for comparison: release cadence, parameter scale, company share, calendar heatmaps, and where each lab sits geographically.
- SwimlanesReleases by company across years
- GrowthCumulative count over time
- DensityQuarterly release heat
- ScaleParameter growth (log)
- LandscapeCompany × date map
- ActivityDaily grid (GitHub-style)
- SharePer-company bars / treemap
- CadenceRelease spacing and bursts
- ChurnNew vs inactive companies by year
- OriginsLabs by region and country
How to explore
guide
$
Narrow it
Filter by open or closed weights, modality, and year. Search matches lab or model names.
$
Follow the source
Use each entry's link to read the primary announcement — the timeline is the index, not the archive.
$
Switch modes
Jump to Graph for aggregates and visual comparisons, or Snapshot for stats, streaks, and calendar oddities — same data, different lenses.
How it's built
Prototype
v0 · VercelI wrote one detailed prompt, dropped it into v0 with Claude Opus 4.6, and it gave me a near-complete working prototype — proper structure, functional layout, the whole thing. From there it was mostly improvements and modifications to get it where I wanted.
Iteration
Cursor · Frontier modelsFrom there I moved into Cursor, picking whichever frontier model made sense for the job at hand. Honestly, most of what you see here was generated or heavily assisted by AI — and that's intentional. Good prompts, the right model, a solid IDE — you can get UI that feels polished and ready to ship. The real work is in steering the thing, not pretending every line was hand-typed from scratch.
Interplay
Amp · CLIAlongside Cursor, I started using Amp CLI for a lot of the work — features, refactors, bug fixes, planning, all of it. I kept switching between the two depending on what felt right for the task. Having a terminal-first agent in the mix made it easy to jump in, steer quick changes, and move on.
Where the data comes from
Collecting the facts
I collected model names, release dates, and relevant links — blog posts, docs, papers, HuggingFace pages, GitHub repos — using a mix of Grok search and Gemini search. They're surprisingly good at digging up accurate release info. The recent ones I mostly already knew. Some came straight from official AI lab pages like API docs and pricing pages.
Writing descriptions
For each model's description, I gave Claude and Gemini focused instructions along with the context I'd gathered — the name, release date, all the links. They'd pull from those sources via web search and write up a description. It's all done manually at the moment, one model at a time.
What's next
This whole flow could be an AI agent with some kind of review/verify step — and that's the plan for future releases so new models can be added automatically. For now, it's manual but it works.
Similar resources
“Other catalogs exist — some narrower, some broader, some optimized for benchmarks rather than narrative history. This project sits in the middle: dense enough to browse, shallow enough per row that you can still skim.”
A Tree of AI Model Names
An interactive tree view of AI model naming across labs — exposing every skipped version, weird suffix, and rebrand in a collapsible file-tree structure.
Models Table
A massive spreadsheet-style catalog of 1000+ LLMs with parameter counts, benchmark scores, training data, and release dates — maintained by Dr Alan D. Thompson.
AI LLM Timeline
A similar chronological timeline of LLM releases from the Transformer era to today — fully AI-agent generated, covers the basics but lacks graphs and deeper data.
AI Timeline - The Road to AGI
A curated timeline spanning 2015–2026, telling the story of the last decade in AI — from cultural trends to technical advancements, with each event linking to source material.