Claude Code with Ollama: Qwen 3.5 setup and Kimi cloud

This tutorial connects Claude Code to Ollama, using Qwen 3.5 on the speaker's Linux laptop before trying Kimi K2.5 through Ollama Cloud. It covers installation and practical coding tasks, with brief installation guidance for macOS and Windows.

The local LLM setup starts with installing Ollama and Claude Code, checking that the server runs, and downloading Qwen 3.5. The speaker creates a custom model with a 64K context window, then launches Claude Code with that model inside a workspace. He presents 64K as Claude Code's minimum context requirement. The examples ask the coding assistant to create and test a C program, write a folder-creation shell script, and build an HTML and CSS page through the VS Code terminal.

Hardware limits matter in this demonstration. The speaker recommends at least 16 GB of RAM and an Nvidia GPU with 4 GB of VRAM, and reports slower responses on his laptop with partial GPU acceleration. These are his stated requirements and observations, rather than a universal guarantee that every model or context size will fit.

The final section uses an Ollama account to connect a device and access Kimi K2.5 in the cloud. The speaker describes a free tier and reports a to-do app taking 30 to 40 seconds. This route uses remote compute rather than the viewer's hardware.