Llama 3.2 comes to Ollama with small local text models and vision models on the way

The source states that Meta's Llama 3.2 is now available to run using Ollama. The release covers four sizes: 1B, 3B, 11B and 90B. The 1B and 3B models are text-only and optimized to run locally on a mobile or edge device. The 11B and 90B models are vision models that support image reasoning use cases, such as document-level understanding including charts and graphs and captioning of images. The source marks the vision models as coming very soon.

For practical everyday work, the smaller models are the part with a clear consequence. The source says they can be used to build highly personalized, on-device agents. It gives examples: asking the model to summarize the last ten messages received on WhatsApp, or to summarize a schedule for the next month. The source says prompts and responses should feel instantaneous, and that with Ollama processing is done locally, maintaining privacy by not sending data such as messages and other information to other third parties or cloud services.

What this does that a previous option could not is not spelled out in the evidence. The source does not compare Llama 3.2 with earlier models or with other local options. The practical change it does describe is the availability of small text-only models intended for local mobile or edge use, plus vision models for image reasoning that are not yet available in the same way. For readers, the useful question is whether local, private summarization and on-device agent tasks become easier to run without sending data to cloud services. The source supports that direction but does not provide benchmarks, device requirements, or a direct before-and-after comparison. That is where the evidence is thin.

If you already use Ollama, the source provides commands: ollama run llama3.2:1b for 1B and ollama run llama3.2 for 3B. It also says to download Ollama and run Llama 3.2. The source does not state pricing or hardware requirements. The vision models are not presented as available now; they are marked coming very soon. That qualifier matters for anyone planning work around image reasoning today.