Ollama gained traction by being the first easy llama.cpp wrapper, then spent years dodging attribution, misleading users, and pivoting to cloud, all while riding VC money earned on someone else's engine. Here's the full history, and why the alternatives are better.

Hacker News is a community-driven platform for sharing and discussing technology news, startups, and programming-related topics. Through user submissions and comments, Hacker News offers insights into emerging technology trends, industry developments, and entrepreneurial ventures. Readers can participate in discussions, share their insights, and stay informed about the latest advancements in technology and innovation.

Hacker News

A detailed critique of Ollama, the popular local LLM runner, arguing it should be abandoned in favor of alternatives. The post documents Ollama's history of failing to credit llama.cpp (its core inference engine), MIT license non-compliance for over 400 days, misleading model naming (e.g., labeling distilled DeepSeek-R1 variants as the full model), shipping a closed-source GUI app, a poorly executed fork of llama.cpp that introduced regressions and 1.8x performance gaps, a proprietary model registry that creates vendor lock-in, and a pivot toward cloud-hosted models with unclear privacy guarantees. The author recommends llama.cpp directly, LM Studio, Jan, llama-swap, LiteLLM, koboldcpp, or Red Hat's ramalama as better alternatives.

Friends Don't Let Friends Use Ollama

The Modelfile: Reinventing a Solved Problem