Ollama v0.32.12 adds Qwen 3.8 27B with Apple Silicon MLX optimization
Run Qwen 3.8 27B locally via Ollama with optimized MLX for Macs
Ollama's latest release, v0.32.12, brings official support for Qwen 3.8 27B, Alibaba's newest open-weight model. The model is positioned as a major upgrade across coding, professional work, research, and long-horizon agentic tasks, making it a strong contender for local deployment. The release landed just hours after the model's debut, reflecting Ollama's fast-moving integration pipeline. For developers, this means they can now pull and run the 27B-parameter model with a single command: `ollama run qwen3.8:27b`.
What's particularly notable is Ollama's optimization for Apple Silicon. The release includes a special MLX variant (`qwen3.8:27b-mlx`) built for maximum performance on Macs, targeting repeated tasks and coding agents where latency and consistency matter. MLX is Apple's array-based machine learning framework, allowing the model to exploit the unified memory architecture of M-series chips. This makes Qwen 3.8 27B more accessible to Mac users who previously struggled with large models on consumer hardware. With Ollama's massive community—179k stars and 17.4k forks on GitHub—this integration is likely to accelerate the model's adoption in local AI workflows and agent-based development.
- Ollama v0.32.12 adds support for Alibaba's Qwen 3.8 27B, a 27B-parameter open-weight model
- Includes a dedicated MLX build for Apple Silicon, optimizing performance on M-series Macs
- Run it with `ollama run qwen3.8:27b` or `ollama run qwen3.8:27b-mlx` for local execution
Why It Matters
Developers can now run a top-tier 27B model locally on Macs, enabling private, fast agentic workflows.