What models power AI proposal tools (Claude, GPT, Llama)?

Mostly Anthropic's Claude (Opus, Sonnet, Haiku) and OpenAI's GPT-4 family, with Llama 3.1 405B and Mistral Large for self-hosted regulated deployments. Well-built tools route by task — Opus for template generation, Sonnet or GPT-4o for fills and rewrites — because using one model for everything either overpays or underdelivers.