Ollama v0.30.8 Released with Prompt Caching Improvements and MLX Inference Stability

Core summary

One sentence to understand this update

Ollama v0.30.8 fixes launch provider selection, improves prompt caching by decoupling it from context shifts for better KV cache reuse, and enhances MLX inference stability.

Impact & opportunity

What this could mean

Builders using Ollama can expect more reliable model launches, improved performance from better prompt caching, and enhanced stability for MLX inference.