NeuralStackIn a fast-moving AI landscape where new models seem to land every week, a small data project has quietly become one of the clearest windows into how quickly the field is evolving. AI Release Tracker, a free, public timeline of major AI model launches, now charts 241 frontier models from 11 leading labs, beginning with the release of ChatGPT on November 30 2022 and continuing through the latest models in 2026. The site has already been cited by Business Insider for its data showing that the monthly cadence of significant AI releases has roughly quadrupled since 2023.
At its core, AI Release Tracker is built around a simple premise: treat every major model launch as a discrete, timestamped event, and make those events searchable and comparable across companies. The homepage renders that idea as a dense horizontal timeline running from late 2022 through 2026, with rows for OpenAI, Anthropic, Google, Meta, SpaceXAI, DeepSeek, Mistral, Moonshot AI, Z.ai, Qwen, and NVIDIA. Each row lists the labs’ successive models, from widely known systems like GPT-4, Claude 3, Gemini, and LLaMA to more recent iterations whose names have only just begun circulating among practitioners.
The tracker goes beyond simple naming and dating. Every model page includes the release date, parameter count, context window, model type, and benchmark scores when those metrics are available. Among the benchmarks tracked, GPQA Diamond focuses on graduate-level science reasoning, while SWE-Bench Verified measures performance on real-world software engineering tasks, and MMMU covers multimodal understanding. An analytics section aggregates this data into charts that show cumulative releases, monthly release frequency, year-over-year totals, benchmark frontier progression, and how quickly each company is shipping new models.
The timeline itself reads like a compressed history of the modern generative AI era. OpenAI’s line marches from GPT-3.5 and GPT-4 through a sprawling GPT-5 family and newer GPT-4.x variants, reflecting rapid iteration on both general and specialized capabilities. Anthropic’s track traces Claude from its first version through Claude 3 and a series of Claude Opus and Sonnet releases, culminating in Claude Opus 5 and other 5-series models. Google’s row charts the evolution from Bard to multiple generations of Gemini and Gemma, while Meta’s entry follows LLaMA from its first release through successive LLaMA 4 variants and related systems such as Code Llama and Muse.
Smaller but increasingly influential players also have fully fleshed-out timelines. SpaceXAI’s Grok series appears with multiple numbered iterations and specialized variants, reflecting the company’s push into chat and coding models. DeepSeek’s line runs from its early coder and LLM releases through several generations of DeepSeek V and R models, including preview and revised editions, culminating in V4 families with Pro and Flash variants. Mistral’s track starts with compact open-weight models and expands into larger Mixtral and Codestral systems and multiple generations of Mistral Small, Medium, and Large, while Qwen and Z.ai each show steady progressions of their own model families.
The tracker also highlights moments when specific models set new high-water marks on public benchmarks. On GPQA Diamond, AI Release Tracker lists GPT-5.4-Pro from OpenAI as the current leader in graduate-level science reasoning. On SWE-Bench Verified, the top slot belongs to Anthropic’s Claude Opus 4.7, which the tracker identifies as the best-performing system on its real-world software engineering tasks. These benchmark leaders are surfaced on the Analytics page alongside historical curves, making it easier to see not just how often models are released but how quickly their frontier performance is moving.
The cadence of new launches has not slowed in 2026. As of late August 2026, AI Release Tracker identifies GLM-5.3-Flash by Z.ai as the most recent frontier model it has recorded, with an August 26 release date. The GLM series sits within a broader GLM family that the site traces from earlier GLM-4 variants through multiple GLM-5 point releases and flash models, illustrating how one lab’s roadmap plays out over months and years. Because the tracker is continuously updated, its homepage doubles as a live feed of which labs are currently shipping and which families are in active development.
Why this matters
The sheer volume and diversity of models in AI Release Tracker’s database underscore how difficult it has become to keep track of the frontier without dedicated tooling. For engineers, researchers, and policymakers, understanding not just which models exist but when they launched, how they differ, and how they perform on standardized benchmarks is increasingly central to decisions about deployment, safety, and regulation. A timeline that spans multiple labs and surfaces both release cadence and benchmark leaders provides a more neutral view of competition and capability progress than any single company’s marketing materials.
AI Release Tracker also reflects a broader shift toward treating model releases as an observable, measurable signal of the industry’s health and direction. By exposing year-over-year totals and monthly release frequency, the analytics page lets observers see how quickly the frontier is advancing and whether that pace is accelerating or stabilizing. The company-level breakdowns show which labs are driving most of the activity, while the benchmark frontier progression highlights that progress is not just about quantity of models but about quality of reasoning, coding, and multimodal understanding.
Looking ahead, the value of a live, cross-lab release timeline is likely to grow as more companies join the race and existing players fragment their offerings into specialized variants. With new GPT, Claude, Gemini, LLaMA, Grok, DeepSeek, Mistral, Qwen, GLM, and related models appearing on an increasingly tight schedule, having a single place that records each launch, its basic specs, and its benchmark performance can help ground debates about AI progress in shared data rather than hype. If the release cadence continues to climb, tools like AI Release Tracker may become as important to following AI as package registries are to software or financial terminals are to markets.