ollama

ollama/ollama

Ollama: Local Open Model Runnerready

Ollama is a tool that provides a CLI and REST API for running and managing open models locally. It is installed and used on macOS, Windows, Linux, and Docker environments, primarily utilized by developers and teams building AI application and agent integrations. Releases are important as they directly impact installation methods, CLI/API compatibility, the model execution experience, and official integrations and platform support.

ailocal-llmcliapiopen-source

Choose version

0.34.21w ago

ollama/ollama v0.34.2

Original

Key point

ollama v0.34.2 adds the ability to choose between account linking or local usage on first run, and fixes a memory leak issue during MLX speculative decoding.

Details

Key Features

First-run setup: Introduced an onboarding process that allows users to choose between logging in with an account or using locally when running ollama. On macOS and Windows, the desktop app and setup completion are synchronized.
Deep link support: Added the ollama://apps URI scheme to directly open the Apps page of the desktop app on macOS and Windows.

Bug Fixes and Performance Improvements

Memory leak fix: Fixed the issue of excessive memory growth during long generation tasks when using MLX speculative decoding.
Engine update: Updated the llama.cpp library to the latest version.

ollama releases | AI Briefing