Open Weights Models Hall of Fame

A Reddit user has sparked a viral discussion by proposing an 'Open Weights Hall of Fame' – a community-curated list of people, companies, and models that have advanced open-weight AI. The post highlights Meta for the entire LLaMA lineage up to 3.3, Mistral for Mixtral 8x7B, Mistral Large, and Mistral Medium 3.5, and OpenAI for Whisper, GPT-OSS-20B/120B, and contributions that formed the foundation for Chinese open-weight models. Google’s Gemma models are recognized for their focus on medical imaging and other niche applications. DeepSeek is credited for DeepSeek-V2/V3/R1 and V4, while Alibaba’s Qwen family (especially Qwen2.5-32B Coder and Qwen3.x) gets a nod for strong performance across coding and reasoning tasks.

Beyond companies, the post honors individuals and community efforts: Georgi Gerganov and the llama.cpp team (including ikrakow and others), quant creators like TheBloke, bartowski, unsloth, and mradermacher who made models accessible to consumer hardware. HuggingFace is thanked for hosting petabytes of model data. The foundational 'Attention is all you need' paper authors and RAG researchers are also enshrined. Honorable mentions include MoonshotAI’s Kimi 2.x, Z-AI’s GLM models, the MLX community for Mac LLM performance, Minimax for coding alternatives, LMStudio for local inference, and Open WebUI for democratizing LLM administration. The list serves as a reminder of the collaborative ecosystem powering the open AI revolution.