Can I run AI locally?

Meta's versatile 8B — great quality/speed ratio
Microsoft's reasoning-focused model
OpenAI's open-weight MoE with configurable reasoning
Multimodal Mistral with vision support
Google's flagship Gemma 3 model
Best open-source coding model at release
R1 reasoning distilled into Qwen 32B — sweet spot
MoE with 16 experts, 17B active params
OpenAI's flagship open-weight MoE — 52.6% SWE-bench
Dense 123B coding model — 72.2% SWE-bench Verified
Massive MoE reasoning model — 37B active
State-of-the-art MoE — 37B active params
1T-param MoE with 384 experts — 32B active, strong agentic coding
Ultra-tiny model for embedded and edge
Meta's smallest Llama for edge devices
Google's tiny Gemma for on-device
Ultralight model for constrained devices
Tiny reasoning model distilled from R1
Lightweight Llama for mobile and edge
Lightweight multilingual reasoning
Microsoft's efficient small model with long context
Compact Qwen 3 for general tasks
Multimodal Gemma with 128K context
High-quality 7B with sliding window attention
Strong multilingual and coding capabilities
R1 reasoning distilled into Qwen 7B
Qwen 3 with thinking mode support
Google's best mid-size open model
Multilingual model supporting 26 languages with 128K context
Hybrid Mamba2 architecture for reasoning
Multimodal vision and text model
Multilingual 12B with 128K context
Excellent quality for its size class
Strong all-rounder with thinking mode
R1 reasoning distilled into Qwen 14B
Hybrid MoE with convolution+attention layers — 2.3B active
Coding-focused model with 256K context — 68% SWE-bench
Flagship native multimodal Qwen 3.5
MoE with only 3.3B active — extremely efficient
MoE with 1M context and 3B active
High-quality reasoning and multilingual
Fully open research model by Allen AI
Optimized for retrieval-augmented generation
Efficient multimodal MoE with 3B active
Large multimodal MoE with 10B active
Large MoE with 39B active params
Massive MoE with 22B active — frontier quality
Largest multimodal Qwen 3.5 MoE
Multimodal MoE with 128 experts — 17B active, 1M context
Largest open-weight dense model by Meta
Largest open coding MoE — 35B active
Improved V3 with hybrid thinking and tool use