Run Local
Find open models that fit your hardware. Filter by tool use support, context window, and get the install command for Ollama or LM Studio.
Updated March 22, 2026 · New models are released constantly — contributions welcome
Your Hardware
GPU VRAM
select to filter modelsSystem RAM
used when running without a GPULlama 3.2 1B
Meta · 1B
Lightest Llama. Runs on anything.
ollama run llama3.2:1bLlama 3.2 3B
Tool useMeta · 3B
Fast and capable for everyday tasks.
ollama run llama3.2:3bQwen 2.5 0.5B
Alibaba · 0.5B
Tiny but surprisingly capable.
ollama run qwen2.5:0.5bQwen 2.5 3B
Tool useAlibaba · 3B
Strong multilingual support.
ollama run qwen2.5:3bGemma 3 1B
Google · 1B
Google's lightest model. Runs on anything.
ollama run gemma3:1bPhi-3 Mini
Microsoft · 3.8B
Punches above its weight. Great for coding.
ollama run phi3:miniGemma 3 4B
Tool useGoogle · 4B
Strong at 4B. Solid tool use and multilingual support.
ollama run gemma3:4bLlama 3.1 8B
Tool useMeta · 8B
Best quality/size ratio for most use cases.
ollama run llama3.1:8bQwen 2.5 7B
Tool useAlibaba · 7B
Excellent code and tool use at 7B scale.
ollama run qwen2.5:7bMistral 7B
Mistral · 7B
Fast, efficient, great instruction following.
ollama run mistral:7bDeepSeek R1 7B
DeepSeek · 7B
Reasoning model. Chain-of-thought distilled.
ollama run deepseek-r1:7bCodeLlama 7B
Meta · 7B
Solid general code generation.
ollama run codellama:7bLLaVA 7B
LLaVA Team · 7B
Vision + language. Understands images.
ollama run llava:7bLlama 3.3 70B
Tool useMeta · 70B
Better than 3.1 70B with the same VRAM. Meta's best open 70B.
ollama run llama3.3:70bLlama 3.1 70B
Tool useMeta · 70B
Near-frontier quality locally.
ollama run llama3.1:70bLlama 3.1 405B
Tool useMeta · 405B
Largest open model. Needs serious hardware.
ollama run llama3.1:405bQwen 2.5 14B
Tool useAlibaba · 14B
Strong all-rounder, great for coding.
ollama run qwen2.5:14bQwen 2.5 32B
Tool useAlibaba · 32B
Top open-source quality below 70B.
ollama run qwen2.5:32bQwen 2.5 72B
Tool useAlibaba · 72B
Flagship Qwen. Best open multilingual model.
ollama run qwen2.5:72bMixtral 8×7B
Mistral · 47B MoE
MoE model with 12.9B active params.
ollama run mixtral:8x7bMistral Small 22B
Tool useMistral · 22B
Best small Mistral for coding and agents.
ollama run mistral-smallMistral Small 3
Tool useMistral · 24B
Latest Mistral Small. Faster and more accurate than its predecessor.
ollama run mistral-small3Phi-3 Medium
Microsoft · 14B
Strong reasoning for a 14B model.
ollama run phi3:mediumPhi-4
Tool useMicrosoft · 14B
Latest Phi. Excellent at STEM and tool use.
ollama run phi4Gemma 3 12B
Tool useGoogle · 12B
Best Gemma 3 for everyday tasks. Great instruction following.
ollama run gemma3:12bGemma 3 27B
Tool useGoogle · 27B
Google's flagship open model. Rivals much larger models.
ollama run gemma3:27bDeepSeek V3
Tool useDeepSeek · 236B MoE
DeepSeek's flagship general model. Rivals GPT-4 class. Needs serious hardware.
ollama run deepseek-v3DeepSeek R1 14B
DeepSeek · 14B
Strong reasoning at 14B.
ollama run deepseek-r1:14bDeepSeek R1 32B
DeepSeek · 32B
Top open reasoning model.
ollama run deepseek-r1:32bDeepSeek Coder V2 16B
DeepSeek · 16B MoE
Best open coding model.
ollama run deepseek-coder-v2:16bCodeLlama 34B
Meta · 34B
Best CodeLlama variant.
ollama run codellama:34bLlama 3.2 Vision 11B
Tool useMeta · 11B
Meta's multimodal model with tool calling.
ollama run llama3.2-vision:11b