Interactive Timeline Reveals Explosive Growth of 171 Large Language Models (2017–2026)

A newly launched interactive timeline, created by independent researcher asymortenson, offers the most comprehensive visual record to date of the rapid evolution of large language models (LLMs) from the advent of the Transformer architecture in 2017 through projected releases up to 2026. The timeline, hosted at llm-timeline.com, catalogs 171 major LLMs developed by 54 organizations worldwide, allowing users to filter by open/closed source, geographic origin, and key milestones. The data reveals a startling acceleration in innovation: over half of all recorded models — 108 in total — were released in just two years, between 2024 and 2025.

The project underscores a pivotal shift in the AI landscape: open-source models achieved parity with proprietary ones in 2025, with 29 open-source and 28 closed-source models released that year. This marks a turning point from the early dominance of closed ecosystems like OpenAI’s GPT series and Google’s PaLM, toward a more democratized, community-driven model development environment. Notably, Chinese research institutions and companies account for approximately 20% of all major releases, with 10 organizations contributing 32 models — a figure that reflects the country’s strategic investment in AI infrastructure and its growing role as a global leader in foundational AI research.

The timeline includes landmark models such as the original Transformer (2017), BERT (2018), GPT-3 (2020), LLaMA (2023), and projected releases like GPT-5.3 Codex, with each entry annotated with release dates, model size, training data sources, and licensing information. Users can search for specific organizations — including Meta, Anthropic, Alibaba, and Mistral AI — or filter by language support, multimodal capabilities, and inference efficiency. The tool also highlights regulatory and ethical milestones, such as the EU AI Act’s impact on model transparency and the release of open-weight models in response to growing public demand for accountability.

While commercial entities like OpenAI and Google still dominate headlines, the data suggests a more fragmented, competitive ecosystem. Academic labs and independent researchers, often leveraging open-source foundations like LLaMA and Mistral, are increasingly responsible for breakthroughs in efficiency and specialization. The rise of models such as Qwen, Baichuan, and DeepSeek from Chinese labs further illustrates the global dispersion of AI innovation, challenging the U.S.-centric narrative that has dominated media coverage.

As the timeline extends into 2026, it includes speculative but plausible releases based on research pipelines, hiring trends, and patent filings. The creator invites the community to contribute missing models, emphasizing transparency and collaborative curation. "This isn’t just a historical record — it’s a living document," asymortenson wrote in the project’s description. "The pace of change is too fast for any single entity to track alone."

Industry analysts caution that while open-source proliferation increases accessibility, it also raises concerns around misuse, model watermarking, and the environmental cost of training increasingly large models. Still, the timeline serves as a vital resource for policymakers, educators, and developers seeking to understand the architecture of modern AI’s explosive growth.

For those navigating the labyrinth of AI development, llm-timeline.com offers an unprecedented lens — not just into the technology, but into the global forces shaping its future.