Closed-source model providers
Anthropic, OpenAI, Google, xAI — the API-only foundation model providers and how they differ in 2026.
Open-weight models
Llama, Mistral, Qwen, DeepSeek, Gemma — when self-hosting open weights makes sense, and when it doesn't.
Inference servers
vLLM, TGI, SGLang, Ollama, llama.cpp — the runtimes that serve open-weight models.
Embedding models
OpenAI, Cohere, Voyage, BGE, E5 — the models that turn text into vectors for retrieval.