Ollama 0.32.0 — Interactive Agent CLI, Web Search, Work Delegation, Gemma 4 90% Faster on Apple Silicon

What's new

• Interactive agent CLI: running ollama without arguments now opens an interactive agent — chat, code generation, web search, real task delegation
• Web Search: built into the CLI agent, no external plugins required
• Gemma 4 MTP: multi-token prediction enabled by default for Gemma 4 on Apple Silicon — ~90% speedup with no configuration changes
• Codex App renamed to ChatGPT: the integrated shortcut is now called ChatGPT
• Deprecated agent models warnings: warnings when using aging models

Why it matters

Ollama is the most widely deployed tool for running LLMs locally (millions of installations). Transforming the CLI into an agent shifts Ollama from a simple model runner to a full-featured local AI assistant. The 90% Gemma 4 speedup is significant — it makes larger local models viable on standard Mac hardware.

How to try it

ollama pull ollama/ollama:0.32.0 or brew upgrade ollama. Run ollama without arguments for interactive mode.