Run local AI models on your PC.

Very light and fast for tiny local tasks.

Ultra-small model for the fastest possible local replies.

Very small Gemma for quick drafts and short replies.

Compact model for lightweight everyday prompts.

Small and sharp for quick local Q&A.

Tiny model optimized for speed and low resource usage.

Ultra-light model for instant local responses.

Small and responsive model for lightweight prompts.

Best lightweight default for daily local chat.

Code model tuned for quick developer assistance.

Microsoft's compact model for efficient local use.

Tiny and efficient model for quick local responses.

A fast, newer everyday model for chat and reasoning.

Strong local quality with good speed and vision support.

Strong bilingual model with good performance.

Specialized coding model for programming assistance.

Nice for coding and structured output.

Great all-rounder if you have enough RAM.

Llama fine-tuned for code generation and understanding.

Smaller instruction model for fast everyday use.

Instruction-tuned model with a chatty, helpful style.

Good general-purpose model with a playful assistant tone.

Friendly instruction model that keeps answers concise.

Coding model for writing and refactoring code.

Strong instruction model for structured responses.

Older but solid general-purpose model.

Bigger StableLM for more capable local chat.

Polished instruction model with strong general responses.

Great all-around local model with strong reasoning.

Improved Llama 3 with better instruction following.

Fine-tuned for instruction following and casual chat.

Google's improved Gemma with better reasoning.

Popular vision model — may be buggy in some cases.

Heavier reasoning model for stronger Macs.

Vision model that can describe and reason about images.

Vision-capable Llama for image understanding and chat.

Larger Gemma variant for better quality and deeper reasoning.

Newer Mistral variant with a strong balance of speed and quality.

A stronger compact model for reasoning and structured tasks.

Dedicated code model for editing, generation, and explanation.

Stronger coding model for larger local development tasks.

A larger general-purpose model for powerful Macs.

Code-focused model for more advanced software work.

Strong larger model for higher-quality local responses.

Heavy Falcon variant for powerful machines.

Powerful large model for demanding workloads.

Very large Llama variant for serious local rigs.

Mixture of experts model for advanced tasks.

Strong instruction model for long-form and tool-like tasks.

Classic instruction model for broad local use.

Instruction-tuned model with solid general reasoning.

Focused on helpful, thoughtful instruction following.

Multilingual model for better cross-language prompts.