Entdecke die prägenden Veröffentlichungen hinter GPT, Claude, Gemini, Llama, Qwen, DeepSeek, Mistral, Grok, GLM und Kimi – von den frühen Skalierungsansätzen bis hin zu multimodalem Schlussfolgern und KI-Agenten.
Released in November 2022, ChatGPT turned instruction-tuned language models into a mass-market interface. The timeline begins at this inflection point, then follows the model families that shaped the competitive market.
By 2023–2024, Claude, Gemini, GPT-4, Llama, Qwen, DeepSeek and Mistral were iterating in parallel. Vision, audio, tool use and longer context windows moved from experimental features to core model capabilities.
The next shift was not only larger models, but models that spend compute on reasoning, call tools and complete longer workflows. Open-weight releases also made capable local AI practical on workstations and home servers.
Release dateThe month of the first public announcement, preview, API release or weight publication—not a later knowledge-cutoff date.
Open weightModel weights are downloadable under the publisher’s license. This does not automatically mean the model meets every definition of open source.
Local AIA practical local-deployment marker for downloadable models with community runtimes. Hardware requirements vary widely by size and quantization.
This is a curated chronology of major general-purpose model milestones, not every checkpoint or API snapshot. Product names and trademarks belong to their respective owners. Verify current availability, licenses and specifications at the linked official source.
The transformer architecture and GPT series began earlier, but ChatGPT’s November 2022 release marked the moment large language models became a mainstream product category. This view focuses on the accelerated multi-company model race that followed.
Open-weight models provide downloadable parameters, but their training data, source code and license permissions may still be restricted. The page uses “open weight” as the more precise label.
Smaller or quantized variants of Llama, Qwen, DeepSeek, Mistral, GLM and Kimi models are common local-AI choices. Actual performance depends on model size, quantization, RAM or VRAM, context length and the inference runtime.
Major model providers now release updates frequently. Review the dataset monthly and add only meaningful model-family milestones, keeping preview and generally available releases clearly distinguished.
Ihr Warenkorb ist derzeit leer.
Jetzt einkaufen
