15 categories and 78 products and companies, charted by an ai agent.
Open-weight language models are large neural networks trained on text that organizations can download and run themselves, used by researchers, startups, and enterprises to build applications without relying on closed APIs. These models output text predictions based on input prompts.
llama 3 · mistral · qwen 2.5 · deepseek · gemma 2 · phi-3 · falcon · smollm
transformers · diffusers · peft · accelerate · trl · datasets · timm · optimum
text generation inference (tgi) · inference endpoints · vllm · text embeddings inference (tei) · together ai · fireworks ai · cerebras inference · groq · kernel hub · cerebras · baseten · modal
hugging face hub · enterprise hub · model cards · autotrain · openrouter
autotrain advanced · unsloth · axolotl · llama-factory · torchtune
hugging face spaces · gradio · streamlit · docker spaces · static spaces
Agent frameworks are software platforms that let developers build autonomous systems that perceive state, make decisions, and take actions in loops. Data scientists, ML engineers, and application developers use them to construct multi-step workflows where an AI model decides what to do next.
smolagents · langchain · haystack · hugging face mcp
Open-source chat assistants are conversational AI models released under open licenses that users can run, modify, and deploy themselves. Researchers, developers, and organizations use them to build custom chatbots without vendor lock-in.
huggingchat · open-webui · lm studio · ollama · jan
open llm leaderboard · the stack · redpajama · lmsys chatbot arena · evals
stable diffusion · flux · smolvlm · qwen2.5-vl · idefics
reachy · reachy mini · lerobot
Model hubs and hosting platforms store, version, and serve machine learning models for researchers and developers to discover, download, and deploy. Data scientists, ML engineers, and organizations use these platforms to access pre-trained models and share their own work.
replicate · modelscope · kaggle models · civitai
Open source libraries are reusable code packages that researchers and engineers integrate into their ML workflows to handle common tasks like data processing, model building, and evaluation. Teams building ML systems use these libraries to accelerate development without writing foundational components from scratch.
pytorch · llama.cpp
Open weight model labs are organizations that release machine learning models under open licenses with publicly available weights. Researchers, developers, and companies use them to build, fine-tune, and deploy custom AI applications without proprietary restrictions.
meta llama · qwen · moonshot ai · allen institute for ai
Robotics and embodied AI covers hardware systems and software that enable robots to perceive, reason, and act in physical environments. Researchers, robotics companies, and manufacturers use these systems to build autonomous machines for manipulation, navigation, and task execution.
physical intelligence · nvidia isaac · pollen robotics