Labs & Companies
A who's-who of the organizations shipping frontier and open-weight models — what they focus on, and the release that put them on the map.
Frontier labs
GPT series, ChatGPT, and the o-series reasoning models that kicked off the "thinking" era.
Claude family and Claude Code, built on Constitutional AI and a safety-first research agenda.
Gemini, AlphaGo, and AlphaFold — the deepest bench in both language and scientific AI.
The Grok family, developed and shipped fast, with tight integration into X (Twitter).
Open-weight challengers
The Llama family and PyTorch — the biggest single force behind open-weight models rivaling closed frontier ones.
Europe's leading open-weight lab, pairing permissively-licensed models with a competitive closed line.
Frontier-class reasoning models trained at a fraction of the cost of comparable Western models.
One of the most active open-weight release cadences of any lab, spanning tiny to frontier-scale models.
Open image-generation models that put text-to-image in reach of consumer hardware.
The GLM family, shipped open-weight and iterated quickly alongside China's other open labs.
The FLUX family, from the researchers who wrote the original Stable Diffusion, now the open image-generation frontier.
Generative media
Video generation aimed at filmmakers and studios, from the early Gen models to its 2026 line.
The Dream Machine video model, built on the neural capture and Gaussian-splatting tools it started with.
Robotics
Foundation models for robot control, the π (pi) vision-language-action series built on a pretrained VLM.
Founding years and headquarters are approximate and can shift with reorganizations; "defining release" is a subjective pick of the model that best put each lab on the map.