Documentation
From first launch
to full control.
Clear guidance for running local models on Intel Macs with AMD GPUs. Learn chat, images, video, agents, APIs, performance tuning and every benchmark you share.
Guides
Find the answer
without digging.
Each guide is focused, linkable and written around the decisions you make inside ToshLLM.
Getting started
Install the signed ToshLLM app, choose the correct build and start a local model.
Read guide →02Work with contextChat and projects
Organize conversations, apply system prompts, attach documents and media, and preserve or export your work.
Read guide →03Choose what fitsModels and performance
Understand VRAM estimates, quantization, dense and MoE models, context size and multi-GPU configurations.
Read guide →04Create on the GPUImage and video studio
Generate, upscale and queue images, create local video, and manage models and VRAM across AMD GPUs.
Read guide →05Connect local clientsLocal API
Use ToshLLM through OpenAI-compatible and Anthropic-compatible APIs from editors, agents, SDKs and other local applications.
Read guide →06Use your own toolsIntegrations
Connect ToshLLM to Claude Code, Anthropic SDKs, VS Code, Zed, OpenCode, Aider and other local clients.
Read guide →07Extend local modelsAgents and MCP
Use local tools, connect MCP servers, review permissions and isolate agent actions when needed.
Read guide →08Measure consistentlyBenchmark guide
Read prompt and generation measurements, compare configurations and share signed results from the app.
Read guide →09Know what leaves the MacPrivacy and security
Learn what remains local, what benchmark sharing sends and how ToshLLM protects submission identity.
Read guide →10Resolve common issuesTroubleshooting
Fix first launch, CPU compatibility, memory pressure, model loading and local server connection problems.
Read guide →No guide matches that search. Try a feature, setting or workflow.
For teams and individuals
Make ToshLLM work for your team.
Need a tailored deployment, help choosing models, or guidance for a fleet of Intel Macs? Tell us what you are building and we will get back to you personally.
Found a reproducible bug? A public GitHub issue helps everyone follow the fix. Open an issue ↗
hello@toshllm.com