LM Studio Bionic and OpenCode: local coding setup

This beginner tutorial sets up LM Studio Bionic as a local coding assistant, then connects OpenCode to its model server. The speaker explains model selection, GPU offload, reasoning settings and context memory before trying practical projects.

For Windows, he recommends a dedicated GPU and choosing a model whose weights and context fit in GPU memory. He says CPU execution can work but runs slowly. For Macs, he suggests Apple M-series hardware with 16 or 32 GB of memory. The runtime guidance pairs Vulkan with AMD or Intel GPUs and CUDA with Nvidia. Context compaction can keep a session going, but it may discard details.

Bionic creates a Star Trek episode website from a CSV file and images in just under eight minutes, according to the speaker. A follow-up prompt corrects a character label. A native C# Windows application takes about an hour and still needs substantial work.

The OpenCode section enables Bionic's local model API and uses a cloud model to configure the connection before switching to a local LLM. A smaller model gathers city council information into a spreadsheet in about 20 minutes, though some Los Angeles email addresses are wrong. These examples show both useful output and the need to check results. The workflow includes cloud-assisted setup, website access and third-party search, so the full demonstration is not offline.