Ollama

Ollama makes running open-weight models locally as simple as ollama run llama3. Nothing leaves your machine; no API key, no token billing, no usage telemetry sent upstream. The software is MIT-licensed and the model library covers Llama, Mistral, Gemma and dozens of others. $200M from a16z creates obvious investor pressure for a monetisation pivot, but the open-source model means the tool’s value cannot be revoked even if the company changes course—forks will proliferate. Rated stage 1 on current design and structural openness, with the VC note as a live risk flag.