A contributing writer says they moved from ChatGPT to Ollama, a free, open-source tool that runs large language models locally on Linux, macOS, and Windows.
A contributing technology writer has abandoned ChatGPT in favor of Ollama, an installable local AI workflow, arguing the move delivers concrete gains in cost, privacy and environmental footprint. The author reports daily use of Ollama instead of cloud — hosted, for-profit services and frames the change as a practical alternative rather than an experimental hack. For individual developers and small teams, the switch promises tighter data control and lower ongoing fees.
Ollama is a free, open-source application you download and run locally on Linux, macOS or Windows to host large language models on your own machine. It provides a graphical interface on macOS and Windows and a command — line interface on Linux; community — built GUIs such as Alpaca and Msty are also compatible. Users fetch and run models locally rather than calling external APIs, keeping inference and prompts on-device.
The author gives specific hardware guidance: any modern CPU will run Ollama, but a minimum of 16 GB system RAM is advised. For automatic GPU acceleration, an NVIDIA GPU with 8 GB or more of VRAM is recommended, or an Apple Silicon Mac (M1, M2 or M3) with 16 GB or more of unified memory. Midrange machines will work but run models more slowly, and heavy inference can limit multitasking on lower — end systems.
Privacy and governance concerns are central to the argument for local inference. Running models locally means queries are not sent to third — party servers where firms might collect and reuse inputs, the writer says. The article also cites an April 2025 legal development in which Ziff Davis filed a lawsuit alleging that OpenAI used copyrighted material in its training data — a line of events the author uses to explain why some users are opting out of hosted LLM services.
Ollama supports a broad library of downloadable models the writer names, including DeepSeek, Gemma, Qwen, Mistral, GPT-OSS and Llama. Under the author’s setup the app and models are "totally and forever free" to download and run, which changes the cost calculus for individuals and small teams wanting local model access. The piece also warns of trade — offs: local inference reduces reliance on data centers and their energy use but shifts compute, maintenance and scaling burdens onto the user’s hardware, with the clearest performance gains appearing on GPU-accelerated or Apple Silicon machines.
Sources
Replies (0)
No replies in this topic yet.
