Llama 3.1 8B
Llama 3.1 CommunityMeta · 8B · Llama 3.1 Community
Meta's versatile 8B — great quality/speed ratio
4.6
GB 77%
·
128K
ctx
·
~29 tok/s
Tight fit 51/100
Can I Run AI locally?
Find out which AI models your machine can actually run.
GPU | VRAM | BW | RAM | CORES 2,560
Estimates based on browser APIs. Actual specs may vary. WebGPU
12 Runs great 7 Runs well 4 Decent 9 Tight fit 24 Barely runs 21 Too heavy
/
Qwen 3.5 9B
Apache 2.0Alibaba · 9B · Apache 2.0
Multimodal Qwen 3.5 mid-size
5.1
GB 85%
·
32K
ctx
·
~26 tok/s
Tight fit 46/100
Phi-4 14B
MITMicrosoft · 14B · MIT
Microsoft's reasoning-focused model
7.7
GB 128% ↗ RAM
·
16K
ctx
·
~9 tok/s
Barely runs 15/100
Mistral Small 3.1 24B
Apache 2.0Mistral AI · 24B · Apache 2.0
Multimodal Mistral with vision support
12.8
GB 213% ↗ RAM
·
128K
ctx
·
~4 tok/s
Barely runs 10/100
GPT-OSS 20B
Apache 2.0OpenAI · 21B · Apache 2.0
OpenAI's open-weight MoE with configurable reasoning
11.3
GB 188% ↗ RAM
·
128K
ctx
·
~4 tok/s
Barely runs 9/100
Gemma 3 27B
GemmaGoogle · 27B · Gemma
Google's flagship Gemma 3 model
14.3
GB 238% ↗ RAM
·
128K
ctx
·
~3 tok/s
Barely runs 9/100
Qwen 2.5 Coder 32B
Apache 2.0Alibaba · 32B · Apache 2.0
Best open-source coding model at release
16.9
GB 282% ↗ RAM
·
128K
ctx
·
~2 tok/s
Barely runs 8/100
Qwen 3 32B
Apache 2.0Alibaba · 32B · Apache 2.0
Qwen 3 flagship dense model
16.9
GB 282% ↗ RAM
·
128K
ctx
·
~2 tok/s
Barely runs 8/100
DeepSeek R1 Distill 32B
MITDeepSeek · 32B · MIT
R1 reasoning distilled into Qwen 32B — sweet spot
16.9
GB 282% ↗ RAM
·
64K
ctx
·
~2 tok/s
Barely runs 8/100
Llama 3.3 70B
Llama 3.3 CommunityMeta · 70B · Llama 3.3 Community
Best open model at 70B class
36.4
GB 607%
·
128K
ctx
·
0 tok/s
Too heavy 0/100
Llama 4 Scout 17B
Llama 4 CommunityMeta · 109B · Llama 4 Community
MoE with 16 experts, 17B active params
56.3
GB 938%
·
128K
ctx
·
0 tok/s
Too heavy 0/100
GPT-OSS 120B
Apache 2.0OpenAI · 117B · Apache 2.0
OpenAI's flagship open-weight MoE — 52.6% SWE-bench
60.4
GB 1007%
·
128K
ctx
·
0 tok/s
Too heavy 0/100
Devstral 2 123B
MRLMistral AI · 123B · MRL
Dense 123B coding model — 72.2% SWE-bench Verified
63.5
GB 1058%
·
256K
ctx
·
0 tok/s
Too heavy 0/100
DeepSeek R1
MITDeepSeek · 671B · MIT
Massive MoE reasoning model — 37B active
344.2
GB 5737%
·
64K
ctx
·
0 tok/s
Too heavy 0/100
DeepSeek V3.2
MITDeepSeek · 685B · MIT
State-of-the-art MoE — 37B active params
351.4
GB 5857%
·
128K
ctx
·
0 tok/s
Too heavy 0/100
Kimi K2
KimiMoonshot AI · 1T · Kimi
1T-param MoE with 384 experts — 32B active, strong agentic coding
512.7
GB 8545%
·
128K
ctx
·
0 tok/s
Too heavy 0/100
All models
Qwen 3.5 0.8B
Apache 2.0Alibaba · 0.8B · Apache 2.0
Ultra-tiny model for embedded and edge
0.9
GB 15%
·
32K
ctx
·
~149 tok/s
Runs great 92/100
Llama 3.2 1B
Llama 3.2 CommunityMeta · 1B · Llama 3.2 Community
Meta's smallest Llama for edge devices
1
GB 17%
·
128K
ctx
·
~134 tok/s
Runs great 92/100
Gemma 3 1B
GemmaGoogle · 1B · Gemma
Google's tiny Gemma for on-device
1
GB 17%
·
32K
ctx
·
~134 tok/s
Runs great 92/100
TinyLlama 1.1B
Apache 2.0Community · 1.1B · Apache 2.0
Ultralight model for constrained devices
1.1
GB 18%
·
2K
ctx
·
~122 tok/s
Runs great 92/100
Qwen 2.5 Coder 1.5B
Apache 2.0Alibaba · 1.5B · Apache 2.0
Ultra-lightweight coding model
1.3
GB 22%
·
32K
ctx
·
~103 tok/s
Runs great 92/100
DeepSeek R1 1.5B
MITDeepSeek · 1.5B · MIT
Tiny reasoning model distilled from R1
1.3
GB 22%
·
64K
ctx
·
~103 tok/s
Runs great 92/100
Qwen 3 1.7B
Apache 2.0Alibaba · 1.7B · Apache 2.0
Compact multilingual Qwen 3
1.4
GB 23%
·
32K
ctx
·
~96 tok/s
Runs great 92/100
Qwen 3 0.6B
Apache 2.0Alibaba · 0.6B · Apache 2.0
Ultra-light Qwen 3 model for constrained devices
0.8
GB 13%
·
32K
ctx
·
~168 tok/s
Runs great 91/100
Qwen 3.5 2B
Apache 2.0Alibaba · 2B · Apache 2.0
Small multimodal Qwen 3.5
1.5
GB 25%
·
32K
ctx
·
~90 tok/s
Runs great 91/100
Gemma 2 2B
GemmaGoogle · 2B · Gemma
Google's compact open model
1.5
GB 25%
·
8K
ctx
·
~90 tok/s
Runs great 91/100
Llama 3.2 3B
Llama 3.2 CommunityMeta · 3B · Llama 3.2 Community
Lightweight Llama for mobile and edge
2
GB 33%
·
128K
ctx
·
~67 tok/s
Runs great 85/100
SmolLM3 3B
Apache 2.0HuggingFace · 3B · Apache 2.0
Lightweight multilingual reasoning
2
GB 33%
·
128K
ctx
·
~67 tok/s
Runs great 85/100
Phi-3.5 Mini
MITMicrosoft · 3.8B · MIT
Microsoft's efficient small model with long context
2.4
GB 40%
·
128K
ctx
·
~56 tok/s
Runs well 79/100
Phi-4 Mini Reasoning
MITMicrosoft · 3.8B · MIT
Lightweight reasoning model
2.4
GB 40%
·
16K
ctx
·
~56 tok/s
Runs well 79/100
Qwen 3 4B
Apache 2.0Alibaba · 4B · Apache 2.0
Compact Qwen 3 for general tasks
2.5
GB 42%
·
32K
ctx
·
~54 tok/s
Runs well 78/100
Gemma 3 4B
GemmaGoogle · 4B · Gemma
Multimodal Gemma with 128K context
2.5
GB 42%
·
128K
ctx
·
~54 tok/s
Runs well 78/100
Qwen 3.5 4B
Apache 2.0Alibaba · 4B · Apache 2.0
Small multimodal Qwen 3.5
2.5
GB 42%
·
32K
ctx
·
~54 tok/s
Runs well 78/100
Gemma 4 E2B IT
GemmaGoogle · 5B · Gemma
Gemma 4 efficient instruct model (official)
3.1
GB 52%
·
256K
ctx
·
~43 tok/s
Runs well 70/100
Gemma 4 E2B
GemmaGoogle · 5B · Gemma
Gemma 4 efficient base model (official)
3.1
GB 52%
·
256K
ctx
·
~43 tok/s
Runs well 70/100
Mistral 7B v0.3
Apache 2.0Mistral AI · 7B · Apache 2.0
High-quality 7B with sliding window attention
4.1
GB 68%
·
32K
ctx
·
~33 tok/s
Decent 57/100
Qwen 2.5 7B
Apache 2.0Alibaba · 7B · Apache 2.0
Strong multilingual and coding capabilities
4.1
GB 68%
·
128K
ctx
·
~33 tok/s
Decent 57/100
Qwen 2.5 Coder 7B
Apache 2.0Alibaba · 7B · Apache 2.0
Dedicated coding model
4.1
GB 68%
·
128K
ctx
·
~33 tok/s
Decent 57/100
DeepSeek R1 Distill 7B
MITDeepSeek · 7B · MIT
R1 reasoning distilled into Qwen 7B
4.1
GB 68%
·
64K
ctx
·
~33 tok/s
Decent 57/100
Gemma 4 E4B IT
GemmaGoogle · 8B · Gemma
Gemma 4 balanced instruct model (official)
4.6
GB 77%
·
256K
ctx
·
~29 tok/s
Tight fit 51/100
Gemma 4 E4B
GemmaGoogle · 8B · Gemma
Gemma 4 balanced base model (official)
4.6
GB 77%
·
256K
ctx
·
~29 tok/s
Tight fit 51/100
Qwen 3 8B
Apache 2.0Alibaba · 8B · Apache 2.0
Qwen 3 with thinking mode support
4.6
GB 77%
·
128K
ctx
·
~29 tok/s
Tight fit 51/100
Ministral 8B
MRLMistral AI · 8B · MRL
Mistral's efficient 8B model
4.6
GB 77%
·
32K
ctx
·
~29 tok/s
Tight fit 51/100
Gemma 2 9B
GemmaGoogle · 9B · Gemma
Google's best mid-size open model
5.1
GB 85%
·
8K
ctx
·
~26 tok/s
Tight fit 46/100
GLM-4 9B
GLM-4Zhipu AI · 9B · GLM-4
Multilingual model supporting 26 languages with 128K context
5.1
GB 85%
·
128K
ctx
·
~26 tok/s
Tight fit 46/100
Nemotron Nano 9B v2
NVIDIA OpenNVIDIA · 9B · NVIDIA Open
Hybrid Mamba2 architecture for reasoning
5.1
GB 85%
·
128K
ctx
·
~26 tok/s
Tight fit 46/100
Llama 3.2 11B Vision
Llama 3.2 CommunityMeta · 11B · Llama 3.2 Community
Multimodal vision and text model
6.1
GB 102%
·
128K
ctx
·
~18 tok/s
Barely runs 26/100
Gemma 3 12B
GemmaGoogle · 12B · Gemma
Multimodal Gemma with 128K context
6.6
GB 110%
·
128K
ctx
·
~14 tok/s
Barely runs 23/100
Mistral Nemo 12B
Apache 2.0Mistral AI · 12B · Apache 2.0
Multilingual 12B with 128K context
6.6
GB 110%
·
128K
ctx
·
~14 tok/s
Barely runs 23/100
Qwen 2.5 14B
Apache 2.0Alibaba · 14B · Apache 2.0
Excellent quality for its size class
7.7
GB 128% ↗ RAM
·
128K
ctx
·
~9 tok/s
Barely runs 15/100
Qwen 3 14B
Apache 2.0Alibaba · 14B · Apache 2.0
Strong all-rounder with thinking mode
7.7
GB 128% ↗ RAM
·
128K
ctx
·
~9 tok/s
Barely runs 15/100
DeepSeek R1 Distill 14B
MITDeepSeek · 14B · MIT
R1 reasoning distilled into Qwen 14B
7.7
GB 128% ↗ RAM
·
64K
ctx
·
~9 tok/s
Barely runs 15/100
LFM2 24B
Liquid AILiquid AI · 24B · Liquid AI
Hybrid MoE with convolution+attention layers — 2.3B active
12.8
GB 213% ↗ RAM
·
32K
ctx
·
~4 tok/s
Barely runs 10/100
Devstral Small 2 24B
Apache 2.0Mistral AI · 24B · Apache 2.0
Coding-focused model with 256K context — 68% SWE-bench
12.8
GB 213% ↗ RAM
·
256K
ctx
·
~4 tok/s
Barely runs 10/100
Gemma 2 27B
GemmaGoogle · 27B · Gemma
Google's largest Gemma 2 model
14.3
GB 238% ↗ RAM
·
8K
ctx
·
~3 tok/s
Barely runs 9/100
Gemma 4 26B-A4B IT
GemmaGoogle · 27B · Gemma
Gemma 4 MoE instruct model (official)
14.3
GB 238% ↗ RAM
·
256K
ctx
·
~3 tok/s
Barely runs 9/100
Gemma 4 26B-A4B
GemmaGoogle · 27B · Gemma
Gemma 4 MoE base model (official)
14.3
GB 238% ↗ RAM
·
256K
ctx
·
~3 tok/s
Barely runs 9/100
Qwen 3.5 27B
Apache 2.0Alibaba · 27.8B · Apache 2.0
Flagship native multimodal Qwen 3.5
14.7
GB 245% ↗ RAM
·
256K
ctx
·
~3 tok/s
Barely runs 9/100
Qwen 3 30B-A3B
Apache 2.0Alibaba · 30B · Apache 2.0
MoE with only 3.3B active — extremely efficient
15.9
GB 265% ↗ RAM
·
128K
ctx
·
~3 tok/s
Barely runs 9/100
Nemotron 3 Nano 30B
NVIDIA OpenNVIDIA · 30B · NVIDIA Open
MoE with 1M context and 3B active
15.9
GB 265% ↗ RAM
·
1024K
ctx
·
~3 tok/s
Barely runs 9/100
Qwen 2.5 32B
Apache 2.0Alibaba · 32B · Apache 2.0
High-quality reasoning and multilingual
16.9
GB 282% ↗ RAM
·
128K
ctx
·
~2 tok/s
Barely runs 8/100
EXAONE 4.0 32B
EXAONE AILG AI · 32B · EXAONE AI
Hybrid reasoning, multilingual
16.9
GB 282% ↗ RAM
·
128K
ctx
·
~2 tok/s
Barely runs 8/100
OLMo 2 32B
Apache 2.0Allen AI · 32B · Apache 2.0
Fully open research model by Allen AI
16.9
GB 282% ↗ RAM
·
4K
ctx
·
~2 tok/s
Barely runs 8/100
Gemma 4 31B IT
GemmaGoogle · 33B · Gemma
Gemma 4 flagship instruct model (official)
17.4
GB 290%
·
256K
ctx
·
0 tok/s
Too heavy 0/100
Gemma 4 31B
GemmaGoogle · 33B · Gemma
Gemma 4 flagship base model (official)
17.4
GB 290%
·
256K
ctx
·
0 tok/s
Too heavy 0/100
Command R 35B
CC BY-NC 4.0Cohere · 35B · CC BY-NC 4.0
Optimized for retrieval-augmented generation
18.4
GB 307%
·
128K
ctx
·
0 tok/s
Too heavy 0/100
Qwen 3.5 35B-A3B
Apache 2.0Alibaba · 35B · Apache 2.0
Efficient multimodal MoE with 3B active
18.4
GB 307%
·
256K
ctx
·
0 tok/s
Too heavy 0/100
Mixtral 8x7B
Apache 2.0Mistral AI · 47B · Apache 2.0
MoE with 12.9B active params
24.6
GB 410%
·
32K
ctx
·
0 tok/s
Too heavy 0/100
Qwen 2.5 72B
QwenAlibaba · 72B · Qwen
Alibaba's flagship open model
37.4
GB 623%
·
128K
ctx
·
0 tok/s
Too heavy 0/100
Qwen 3.5 122B-A10B
Apache 2.0Alibaba · 122B · Apache 2.0
Large multimodal MoE
63
GB 1050%
·
256K
ctx
·
0 tok/s
Too heavy 0/100
Mixtral 8x22B
Apache 2.0Mistral AI · 141B · Apache 2.0
Large MoE with 39B active params
72.7
GB 1212%
·
64K
ctx
·
0 tok/s
Too heavy 0/100
Qwen 3 235B-A22B
Apache 2.0Alibaba · 235B · Apache 2.0
Massive MoE with 22B active — frontier quality
120.9
GB 2015%
·
128K
ctx
·
0 tok/s
Too heavy 0/100
Qwen 3.5 397B-A17B
Apache 2.0Alibaba · 397B · Apache 2.0
Largest multimodal Qwen 3.5 MoE
203.9
GB 3398%
·
256K
ctx
·
0 tok/s
Too heavy 0/100
Llama 4 Maverick 17B-128E
Llama 4 CommunityMeta · 400B · Llama 4 Community
Multimodal MoE with 128 experts — 17B active, 1M context
205.4
GB 3423%
·
1024K
ctx
·
0 tok/s
Too heavy 0/100
Llama 3.1 405B
Llama 3.1 CommunityMeta · 405B · Llama 3.1 Community
Largest open-weight dense model by Meta
208
GB 3467%
·
128K
ctx
·
0 tok/s
Too heavy 0/100
Qwen 3 Coder 480B
Apache 2.0Alibaba · 480B · Apache 2.0
Largest open coding MoE — 35B active
246.4
GB 4107%
·
256K
ctx
·
0 tok/s
Too heavy 0/100
DeepSeek V3.1
MITDeepSeek · 671B · MIT
Improved V3 with hybrid thinking and tool use
344.2
GB 5737%
·
128K
ctx
·
0 tok/s
Too heavy 0/100