AI models directory for capability-aware discovery

Qwen2.5 is a family of large language models by Alibaba Cloud's Qwen team, ranging from 0.5B to 72B parameters, with improved reasoning, coding, multilingual support, and up to 128K token context. Released under Apache 2.0 (except 3B and 72B under Qwen license).

DeepSeek-R1 is a family of open reasoning large language models developed by DeepSeek, available via Ollama for local deployment. It excels in complex reasoning, mathematics, coding, and logical problem-solving, with performance approaching leading models like O3 and Gemini 2.5 Pro. The model is released under the MIT License and can be run locally using `ollama run deepseek-r1`, supporting various sizes from 1.5B to 671B parameters.

The cross-encoder/ms-marco-MiniLM-L6-v2 is a cross-encoder model for text ranking, fine-tuned on the MS MARCO passage ranking dataset. Based on MiniLM with 6 layers, it balances efficiency and accuracy, with over 79.9 million downloads on Hugging Face. It outputs relevance scores for query-passage pairs and supports multiple frameworks including PyTorch, JAX, ONNX, SafeTensors, and OpenVINO. Licensed under Apache-2.0, it is widely used for search re-ranking, question answering, and information retrieval.

Kokoro-82M is an open-weight text-to-speech model with 82 million parameters, fine-tuned from StyleTTS2-LJSpeech. It supports English and Arabic, delivers quality comparable to larger models, and is released under Apache 2.0. With over 16.7 million downloads, it is designed for efficient deployment in production and personal projects.

Llama 3.3 is a state-of-the-art 70B parameter large language model from Meta, optimized for local deployment via Ollama. It offers performance comparable to Llama 3.1 405B, supports up to 128K token context, and is designed for text generation, reasoning, instruction following, and multilingual dialogue.

Gemma 2 is a family of lightweight, state-of-the-art open models from Google, built from the same research and technology used to create the Gemini models. It is available in 2B, 9B, and 27B parameter sizes, designed for efficient deployment on resource-constrained devices and cloud environments. The model excels in text generation, reasoning, and coding tasks, and is optimized for local inference via Ollama.