← CanIRunLocal · All hardware · All models

Can I run it locally?

Pick a model to see which GPUs, Macs, and multi-node clusters can run it, at what quantization and roughly how fast.

Meta

• Llama 3.2 1B — fits M1 8GB
• Llama 3.2 3B — fits M1 8GB
• Llama 3.1 8B — fits RTX 3060 Ti 8GB
• Llama 3.3 70B — fits M1 Max 64GB
• Code Llama 7B — fits M1 8GB
• Llama 3.2 Vision 11B — fits RTX 3060 Ti 8GB
• Llama 4 Scout 109B (MoE) — fits Mac Studio M3 Ultra 96GB
• Llama 4 Maverick 400B (MoE) — fits Mac Studio M3 Ultra 256GB

Alibaba

• Qwen 2.5 3B — fits M1 8GB
• Qwen 2.5 7B — fits M1 8GB
• Qwen 2.5 14B — fits RTX 3060 Ti 8GB
• Qwen 2.5 32B — fits RTX 4060 Ti 16GB
• Qwen 2.5 72B — fits M1 Max 64GB
• Qwen3 0.6B — fits M1 8GB
• Qwen3 1.7B — fits M1 8GB
• Qwen3 4B — fits M1 8GB
• Qwen3 8B — fits RTX 3060 Ti 8GB
• Qwen3 14B — fits RTX 3060 Ti 8GB
• Qwen3 30B-A3B (MoE) — fits RTX 3060 12GB
• Qwen3 32B — fits RTX 4060 Ti 16GB
• Qwen 2.5 Coder 1.5B — fits M1 8GB
• Qwen 2.5 Coder 7B — fits M1 8GB
• Qwen 2.5 Coder 14B — fits RTX 3060 Ti 8GB
• Qwen 2.5 Coder 32B — fits RTX 4060 Ti 16GB
• Qwen 2.5 VL 7B — fits RTX 3060 Ti 8GB
• Qwen3-Coder 30B-A3B (MoE) — fits RTX 3060 12GB
• Qwen3-Coder 480B-A35B (MoE) — fits 4x Mac Studio M5 Max (40c GPU) 128GB
• Qwen3 235B-A22B (MoE) — fits M2 Ultra 192GB
• Qwen3-Next 80B-A3B (MoE) — fits M1 Max 64GB
• Qwen3.8 27B — fits RTX 3080 10GB
• Qwen3.8-Flash-Next 125B-A6B (MoE) — fits M3 Max 128GB

DeepSeek

• DeepSeek Coder V2 Lite 16B (MoE) — fits RTX 3060 Ti 8GB
• DeepSeek R1 Distill Qwen 1.5B — fits M1 8GB
• DeepSeek R1 Distill Qwen 7B — fits M1 8GB
• DeepSeek R1 0528 Qwen3 8B — fits RTX 3060 Ti 8GB
• DeepSeek R1 Distill Qwen 14B — fits RTX 3060 Ti 8GB
• DeepSeek R1 Distill Qwen 32B — fits RTX 4060 Ti 16GB
• DeepSeek R1 Distill Llama 70B — fits M1 Max 64GB
• DeepSeek R1 671B (full, MoE) — fits 2x Mac Studio M5 Ultra (80c GPU) 256GB
• DeepSeek V3.1 671B (MoE) — fits 2x Mac Studio M5 Ultra (80c GPU) 256GB

Mistral

• Mistral 7B v0.3 — fits M1 8GB
• Mistral Nemo 12B — fits RTX 3060 Ti 8GB
• Mistral Small 24B — fits RTX 3060 Ti 8GB
• Mixtral 8x7B (MoE) — fits RTX 3090 24GB

Google

• Gemma 3 1B — fits M1 8GB
• Gemma 3 4B — fits M1 8GB
• Gemma 3 12B — fits RTX 3060 Ti 8GB
• Gemma 3 27B — fits RTX 3080 10GB
• Gemma 3n E2B (on-device) — fits M1 8GB
• Gemma 3n E4B (on-device) — fits RTX 3060 Ti 8GB
• EmbeddingGemma 2 270M — fits M1 8GB
• EmbeddingGemma 2 740M — fits M1 8GB

Microsoft

• Phi-4 Mini 3.8B — fits M1 8GB
• Phi-4 14B — fits RTX 3060 Ti 8GB

OpenAI

• gpt-oss 20B (MoE, MXFP4) — fits RTX 3060 Ti 8GB
• gpt-oss 120B (MoE, MXFP4) — fits Mac Studio M3 Ultra 96GB

LLaVA

• LLaVA 1.6 7B — fits RTX 3060 Ti 8GB

OpenBMB

• MiniCPM-V 2.6 8B — fits RTX 3060 Ti 8GB

Cloudflare

• Clef-Flash 9B (Cloudflare, decision) — fits RTX 3060 Ti 8GB
• Clef 27B (Cloudflare, decision) — fits RTX 3060 12GB

Bespoke Labs

• Nimble v3 9B (Bespoke Labs, decision) — fits RTX 3060 Ti 8GB

Together AI

• tev1 4B (Together AI, decision) — fits M1 8GB

NVIDIA

• Nemotron 3.5 Super 120B-A12B (MoE) — fits M3 Max 128GB
• Nemotron 3.5 Bolt 74B-A9B (MoE, IQ2-class) — fits RTX 3090 24GB

Ornith

• Ornith-1.5 9B — fits RTX 3060 Ti 8GB
• Ornith-1.5 35B-A3B (MoE) — fits RTX 4060 Ti 16GB
• Ornith-1.5 397B (MoE) — fits 4x Mac Studio M5 Max (40c GPU) 128GB

Z.ai

• GLM-5.3-Flash 320B-A18B (MoE) — fits Mac Studio M3 Ultra 256GB

Estimates, not benchmarks. Model sizes are ollama.com download sizes; memory needed adds headroom for ~8k tokens of context; speeds are rough figures from memory bandwidth. Data updated 2026-10-16.

A project by OnDevice Labs · llms.txt