Ollama-compatible models for local PCs

This selection keeps only models that are realistic for a local machine, from an i5 laptop with 16 GB of RAM to a gaming PC with a dedicated GPU.

Quick reading

The listed sizes match the weights published in the Ollama library. Actual memory usage also depends on context size, operating system, number of open conversations and CPU or GPU usage.

16 GB RAM laptop: prefer models up to 8B, ideally between 1B and 7B.
Comfortable 16-32 GB RAM PC: 8B to 14B models become interesting, especially with a GPU.
32-64 GB RAM gaming PC or 16-24 GB VRAM GPU: 27B to 32B models are possible, but slower and more demanding.

Model table

Resource use	Ollama model	Size	Recommended use	Target machine	Command
Very light	nomic-embed-text	274 MB	Embeddings, document search, RAG. Not a chat model.	All PCs	`ollama pull nomic-embed-text`
Very light	smollm2:135m	271 MB	Tests, very simple answers, strong constraints.	All PCs	`ollama run smollm2:135m`
Very light	qwen2.5-coder:0.5b	398 MB	Small coding help, lightweight autocompletion.	All PCs	`ollama run qwen2.5-coder:0.5b`
Very light	qwen3:0.6b	523 MB	Minimal assistant, rephrasing, quick tasks.	All PCs	`ollama run qwen3:0.6b`
Very light	tinyllama	638 MB	Discovering Ollama, quick trials, low expected quality.	All PCs	`ollama run tinyllama`
Light	gemma3:1b	815 MB	Simple questions, short summaries, light multilingual use.	16 GB RAM laptop	`ollama run gemma3:1b`
Light	qwen2.5-coder:1.5b	986 MB	Light coding, script explanations, small fixes.	16 GB RAM laptop	`ollama run qwen2.5-coder:1.5b`
Light	deepseek-r1:1.5b	1.1 GB	Light reasoning, simple math, drafts.	16 GB RAM laptop	`ollama run deepseek-r1:1.5b`
Light	llama3.2:1b	1.3 GB	Very compact general assistant, rephrasing, summarizing.	16 GB RAM laptop	`ollama run llama3.2:1b`
Light	qwen3:1.7b	1.4 GB	Light general assistant, better compromise than sub-1B models.	16 GB RAM laptop	`ollama run qwen3:1.7b`
Light	smollm2:1.7b	1.8 GB	Compact assistant, simple local tasks.	16 GB RAM laptop	`ollama run smollm2`
Standard	qwen2.5-coder:3b	1.9 GB	Light local coding, function review, TypeScript/Python help.	16 GB RAM laptop	`ollama run qwen2.5-coder:3b`
Standard	llama3.2	2.0 GB	General assistant recommended for modest machines.	16 GB RAM laptop	`ollama run llama3.2`
Standard	qwen3:4b	2.5 GB	Very good general compromise for reasoning and writing.	16 GB RAM laptop	`ollama run qwen3:4b`
Standard	gemma3:4b	3.3 GB	General assistant with image support, summarizing, Q&A.	16 GB RAM laptop or light GPU	`ollama run gemma3:4b`
Comfort	mistral	4.4 GB	Fast general assistant, good default 7B choice.	16 GB RAM laptop, better with GPU	`ollama run mistral`
Comfort	qwen2.5-coder	4.7 GB	Versatile local coding, generation and correction.	16 GB RAM laptop, better with GPU	`ollama run qwen2.5-coder`
Comfort	llava:7b	4.7 GB	Image analysis, simple visual questions.	16-32 GB RAM PC or GPU	`ollama run llava`
Comfort	deepseek-r1:7b	4.7 GB	Local reasoning, math, logic, detailed steps.	16-32 GB RAM PC or GPU	`ollama run deepseek-r1:7b`
Comfort	qwen3	5.2 GB	Solid general assistant, reasoning and tool use.	16-32 GB RAM PC or GPU	`ollama run qwen3`
Heavy	gemma3:12b	8.1 GB	Better general quality, vision, long-form writing.	32 GB RAM PC or 12 GB VRAM GPU	`ollama run gemma3:12b`
Heavy	qwen2.5-coder:14b	9.0 GB	More reliable coding, refactoring and complex explanations.	32 GB RAM PC or 12 GB VRAM GPU	`ollama run qwen2.5-coder:14b`
Heavy	deepseek-r1:14b	9.0 GB	More robust reasoning, slower math/code work.	32 GB RAM PC or 12 GB VRAM GPU	`ollama run deepseek-r1:14b`
Heavy	qwen3:14b	9.3 GB	High-quality general assistant with good reasoning level.	32 GB RAM PC or 12 GB VRAM GPU	`ollama run qwen3:14b`
Gaming PC	gemma3:27b	17 GB	High quality, vision, long tasks. Slower without a large GPU.	32-64 GB RAM gaming PC or 24 GB VRAM GPU	`ollama run gemma3:27b`
Gaming PC	qwen3:30b	19 GB	Advanced reasoning, long context, demanding use.	32-64 GB RAM gaming PC or 24 GB VRAM GPU	`ollama run qwen3:30b`
Gaming PC	qwen2.5-coder:32b	20 GB	Advanced coding, multi-file generation, technical reasoning.	32-64 GB RAM gaming PC or 24 GB VRAM GPU	`ollama run qwen2.5-coder:32b`
Gaming PC	qwen3:32b	20 GB	Powerful general assistant, but resource intensive.	32-64 GB RAM gaming PC or 24 GB VRAM GPU	`ollama run qwen3:32b`
Gaming PC	deepseek-r1:32b	20 GB	Advanced reasoning, logic, math, code with longer response times.	32-64 GB RAM gaming PC or 24 GB VRAM GPU	`ollama run deepseek-r1:32b`

Quick recommendations

Best starting point on a 16 GB laptop llama3.2, qwen3:4b or mistral.

Best choice for coding qwen2.5-coder in 7B, then qwen2.5-coder:14b if the machine allows it.

Best choice for reasoning deepseek-r1:7b or deepseek-r1:14b, depending on available resources.

Best choice with images gemma3:4b to start, then gemma3:12b or llava on a more comfortable machine.

Main sources: official Ollama library for available models, Qwen3, Gemma 3, Llama 3.2, DeepSeek-R1, Qwen2.5-Coder, Mistral, LLaVA, SmolLM2, TinyLlama and nomic-embed-text.