https://www.canirun.ai/

CanIRun.ai
[tier list] [docs] [why]

Can I Run AI locally?

Find out which AI models your machine can actually run.

[detecting...] | [--] | [] | WebGPU

Estimates based on browser APIs. Actual specs may vary.

S 0 A 0 B 0 C 0 D 0 F 0
[                    ] /
[All grades     ] [All tasks] [All providers] [Sort: Score   ] List
 

Llama 3.1 8B

1 year ago

Meta * 8B

Meta's versatile 8B -- great quality/speed ratio

4.1 GB * 128K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2024-07
Architecture Dense
Memory --
chatcodereasoning
 

Qwen 3.5 9B

1 month ago

Alibaba * 9B

Multimodal Qwen 3.5 mid-size

4.6 GB * 32K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2026-02
Architecture Dense
Memory --
chatvision
 

Phi-4 14B

1 year ago

Microsoft * 14B

Microsoft's reasoning-focused model

7.2 GB * 16K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2024-12
Architecture Dense
Memory --
reasoningcode
 

GPT-OSS 20B

7mo ago

OpenAI * 21B

OpenAI's open-weight MoE with configurable reasoning

10.8 GB * 128K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2025-08
Architecture MoE
Active 3.6B active
Memory --
chatreasoningcode
 

Mistral Small 3.1 24B

1 year ago

Mistral AI * 24B

Multimodal Mistral with vision support

12.3 GB * 128K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2025-03
Architecture Dense
Memory --
chatvisioncode
 

Gemma 3 27B

1 year ago

Google * 27B

Google's flagship Gemma 3 model

13.8 GB * 128K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2025-03
Architecture Dense
Memory --
chatvisionreasoning
 

Qwen 2.5 Coder 32B

1 year ago

Alibaba * 32B

Best open-source coding model at release

16.4 GB * 128K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2024-11
Architecture Dense
Memory --
code
 

Qwen 3 32B

11mo ago

Alibaba * 32B

Qwen 3 flagship dense model

16.4 GB * 128K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2025-04
Architecture Dense
Memory --
chatcodereasoning
 

DeepSeek R1 Distill 32B

1 year ago

DeepSeek * 32B

R1 reasoning distilled into Qwen 32B -- sweet spot

16.4 GB * 64K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2025-01
Architecture Dense
Memory --
reasoning
 

Llama 3.3 70B

1 year ago

Meta * 70B

Best open model at 70B class

35.9 GB * 128K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2024-12
Architecture Dense
Memory --
chatreasoningcode
 

Llama 4 Scout 17B

11mo ago

Meta * 109B

MoE with 16 experts, 17B active params

55.8 GB * 128K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2025-04
Architecture MoE
Active 17B active
Memory --
chatvisionreasoning
 

GPT-OSS 120B

7mo ago

OpenAI * 117B

OpenAI's flagship open-weight MoE -- 52.6% SWE-bench

59.9 GB * 128K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2025-08
Architecture MoE
Active 5.1B active
Memory --
chatreasoningcode
 

Devstral 2 123B

3mo ago

Mistral AI * 123B

Dense 123B coding model -- 72.2% SWE-bench Verified

63 GB * 256K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2025-12
Architecture Dense
Memory --
code
 

DeepSeek R1

1 year ago

DeepSeek * 671B

Massive MoE reasoning model -- 37B active

343.7 GB * 64K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2025-01
Architecture MoE
Active 37B active
Memory --
reasoning
 

DeepSeek V3.2

3mo ago

DeepSeek * 685B

State-of-the-art MoE -- 37B active params

350.9 GB * 128K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2025-12
Architecture MoE
Active 37B active
Memory --
chatcodereasoning
 

Kimi K2

8mo ago

Moonshot AI * 1T

1T-param MoE with 384 experts -- 32B active, strong agentic coding

512.2 GB * 128K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2025-07
Architecture MoE
Active 32B active
Memory --
chatreasoningcode
All models
 

Qwen 3.5 0.8B

1 month ago

Alibaba * 0.8B

Ultra-tiny model for embedded and edge

0.5 GB * 32K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2026-02
Architecture Dense
Memory --
chatedge
 

Llama 3.2 1B

1 year ago

Meta * 1B

Meta's smallest Llama for edge devices

0.5 GB * 128K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2024-09
Architecture Dense
Memory --
chatedge
 

Gemma 3 1B

1 year ago

Google * 1B

Google's tiny Gemma for on-device

0.5 GB * 32K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2025-03
Architecture Dense
Memory --
chatedge
 

TinyLlama 1.1B

2y ago

Community * 1.1B

Ultralight model for constrained devices

0.6 GB * 2K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2024-01
Architecture Dense
Memory --
chatedge
 

Qwen 2.5 Coder 1.5B

1 year ago

Alibaba * 1.5B

Ultra-lightweight coding model

0.8 GB * 32K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2024-11
Architecture Dense
Memory --
code
 

DeepSeek R1 1.5B

1 year ago

DeepSeek * 1.5B

Tiny reasoning model distilled from R1

0.8 GB * 64K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2025-01
Architecture Dense
Memory --
reasoning
 

Qwen 3 1.7B

11mo ago

Alibaba * 1.7B

Compact multilingual Qwen 3

0.9 GB * 32K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2025-04
Architecture Dense
Memory --
chatmultilingual
 

Qwen 3.5 2B

1 month ago

Alibaba * 2B

Small multimodal Qwen 3.5

1 GB * 32K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2026-02
Architecture Dense
Memory --
chatmultilingual
 

Gemma 2 2B

1 year ago

Google * 2B

Google's compact open model

1 GB * 8K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2024-06
Architecture Dense
Memory --
chatedge
 

Llama 3.2 3B

1 year ago

Meta * 3B

Lightweight Llama for mobile and edge

1.5 GB * 128K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2024-09
Architecture Dense
Memory --
chatcode
 

SmolLM3 3B

8mo ago

HuggingFace * 3B

Lightweight multilingual reasoning

1.5 GB * 128K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2025-07
Architecture Dense
Memory --
chatreasoning
 

Phi-3.5 Mini

1 year ago

Microsoft * 3.8B

Microsoft's efficient small model with long context

1.9 GB * 128K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2024-08
Architecture Dense
Memory --
reasoningcodechat
 

Phi-4 Mini Reasoning

11mo ago

Microsoft * 3.8B

Lightweight reasoning model

1.9 GB * 16K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2025-04
Architecture Dense
Memory --
reasoning
 

Qwen 3 4B

11mo ago

Alibaba * 4B

Compact Qwen 3 for general tasks

2 GB * 32K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2025-04
Architecture Dense
Memory --
chatcode
 

Gemma 3 4B

1 year ago

Google * 4B

Multimodal Gemma with 128K context

2 GB * 128K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2025-03
Architecture Dense
Memory --
chatvision
 

Qwen 3.5 4B

1 month ago

Alibaba * 4B

Small multimodal Qwen 3.5

2 GB * 32K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2026-02
Architecture Dense
Memory --
chatmultilingual
 

Mistral 7B v0.3

1 year ago

Mistral AI * 7B

High-quality 7B with sliding window attention

3.6 GB * 32K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2024-05
Architecture Dense
Memory --
chatreasoning
 

Qwen 2.5 7B

1 year ago

Alibaba * 7B

Strong multilingual and coding capabilities

3.6 GB * 128K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2024-09
Architecture Dense
Memory --
chatmultilingualcode
 

Qwen 2.5 Coder 7B

1 year ago

Alibaba * 7B

Dedicated coding model

3.6 GB * 128K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2024-11
Architecture Dense
Memory --
code
 

DeepSeek R1 Distill 7B

1 year ago

DeepSeek * 7B

R1 reasoning distilled into Qwen 7B

3.6 GB * 64K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2025-01
Architecture Dense
Memory --
reasoning
 

Qwen 3 8B

11mo ago

Alibaba * 8B

Qwen 3 with thinking mode support

4.1 GB * 128K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2025-04
Architecture Dense
Memory --
chatcodereasoning
 

Ministral 8B

1 year ago

Mistral AI * 8B

Mistral's efficient 8B model

4.1 GB * 32K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2024-10
Architecture Dense
Memory --
chat
 

Gemma 2 9B

1 year ago

Google * 9B

Google's best mid-size open model

4.6 GB * 8K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2024-06
Architecture Dense
Memory --
chatreasoning
 

GLM-4 9B

1 year ago

Zhipu AI * 9B

Multilingual model supporting 26 languages with 128K context

4.6 GB * 128K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2024-06
Architecture Dense
Memory --
chatmultilingualcode
 

Nemotron Nano 9B v2

9mo ago

NVIDIA * 9B

Hybrid Mamba2 architecture for reasoning

4.6 GB * 128K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2025-06
Architecture Dense
Memory --
reasoning
 

Llama 3.2 11B Vision

1 year ago

Meta * 11B

Multimodal vision and text model

5.6 GB * 128K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2024-09
Architecture Dense
Memory --
chatvision
 

Gemma 3 12B

1 year ago

Google * 12B

Multimodal Gemma with 128K context

6.1 GB * 128K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2025-03
Architecture Dense
Memory --
chatvisionreasoning
 

Mistral Nemo 12B

1 year ago

Mistral AI * 12B

Multilingual 12B with 128K context

6.1 GB * 128K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2024-07
Architecture Dense
Memory --
chatmultilingual
 

Qwen 2.5 14B

1 year ago

Alibaba * 14B

Excellent quality for its size class

7.2 GB * 128K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2024-09
Architecture Dense
Memory --
chatmultilingualreasoning
 

Qwen 3 14B

11mo ago

Alibaba * 14B

Strong all-rounder with thinking mode

7.2 GB * 128K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2025-04
Architecture Dense
Memory --
chatcodereasoning
 

DeepSeek R1 Distill 14B

1 year ago

DeepSeek * 14B

R1 reasoning distilled into Qwen 14B

7.2 GB * 64K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2025-01
Architecture Dense
Memory --
reasoning
 

LFM2 24B

4mo ago

Liquid AI * 24B

Hybrid MoE with convolution+attention layers -- 2.3B active

12.3 GB * 32K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2025-11
Architecture MoE
Active 2.3B active
Memory --
chatedgerag
 

Devstral Small 2 24B

3mo ago

Mistral AI * 24B

Coding-focused model with 256K context -- 68% SWE-bench

12.3 GB * 256K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2025-12
Architecture Dense
Memory --
code
 

Gemma 2 27B

1 year ago

Google * 27B

Google's largest Gemma 2 model

13.8 GB * 8K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2024-06
Architecture Dense
Memory --
chatreasoning
 

Qwen 3.5 27B

1 month ago

Alibaba * 27.8B

Flagship native multimodal Qwen 3.5

14.2 GB * 256K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2026-02
Architecture Dense
Memory --
chatvisionreasoning
 

Qwen 3 30B-A3B

11mo ago

Alibaba * 30B

MoE with only 3.3B active -- extremely efficient

15.4 GB * 128K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2025-04
Architecture MoE
Active 3.3B active
Memory --
chatreasoning
 

Nemotron 3 Nano 30B

9mo ago

NVIDIA * 30B

MoE with 1M context and 3B active

15.4 GB * 1024K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2025-06
Architecture MoE
Active 3B active
Memory --
chatreasoning
 

Qwen 2.5 32B

1 year ago

Alibaba * 32B

High-quality reasoning and multilingual

16.4 GB * 128K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2024-09
Architecture Dense
Memory --
chatmultilingualreasoning
 

EXAONE 4.0 32B

8mo ago

LG AI * 32B

Hybrid reasoning, multilingual

16.4 GB * 128K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2025-07
Architecture Dense
Memory --
chatreasoning
 

OLMo 2 32B

1 year ago

Allen AI * 32B

Fully open research model by Allen AI

16.4 GB * 4K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2025-03
Architecture Dense
Memory --
chatreasoning
 

Command R 35B

2y ago

Cohere * 35B

Optimized for retrieval-augmented generation

17.9 GB * 128K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2024-03
Architecture Dense
Memory --
chatrag
 

Qwen 3.5 35B-A3B

1 month ago

Alibaba * 35B

Efficient multimodal MoE with 3B active

17.9 GB * 256K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2026-02
Architecture MoE
Active 3B active
Memory --
chatvision
 

Mixtral 8x7B

2y ago

Mistral AI * 47B

MoE with 12.9B active params

24.1 GB * 32K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2023-12
Architecture MoE
Active 12.9B active
Memory --
chatcode
 

Qwen 2.5 72B

1 year ago

Alibaba * 72B

Alibaba's flagship open model

36.9 GB * 128K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2024-09
Architecture Dense
Memory --
chatmultilingualreasoningcode
 

Qwen 3.5 122B-A10B

1 month ago

Alibaba * 122B

Large multimodal MoE with 10B active

62.5 GB * 256K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2026-02
Architecture MoE
Active 10B active
Memory --
chatvisionreasoning
 

Mixtral 8x22B

1 year ago

Mistral AI * 141B

Large MoE with 39B active params

72.2 GB * 64K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2024-04
Architecture MoE
Active 39B active
Memory --
chatcodereasoning
 

Qwen 3 235B-A22B

11mo ago

Alibaba * 235B

Massive MoE with 22B active -- frontier quality

120.4 GB * 128K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2025-04
Architecture MoE
Active 22B active
Memory --
chatcodereasoning
 

Qwen 3.5 397B-A17B

1 month ago

Alibaba * 397B

Largest multimodal Qwen 3.5 MoE

203.4 GB * 256K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2026-02
Architecture MoE
Active 17B active
Memory --
chatvisionreasoningcode
 

Llama 4 Maverick 17B-128E

11mo ago

Meta * 400B

Multimodal MoE with 128 experts -- 17B active, 1M context

204.9 GB * 1024K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2025-04
Architecture MoE
Active 17B active
Memory --
chatvisionreasoningcode
 

Llama 3.1 405B

1 year ago

Meta * 405B

Largest open-weight dense model by Meta

207.5 GB * 128K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2024-07
Architecture Dense
Memory --
chatreasoningcode
 

Qwen 3 Coder 480B

8mo ago

Alibaba * 480B

Largest open coding MoE -- 35B active

245.9 GB * 256K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2025-07
Architecture MoE
Active 35B active
Memory --
code
 

DeepSeek V3.1

7mo ago

DeepSeek * 671B

Improved V3 with hybrid thinking and tool use

343.7 GB * 128K ctx *
Q2_K Q3_K_M Q4_K_M Q5_K_M Q6_K Q8_0 F16
Released 2025-08
Architecture MoE
Active 37B active
Memory --
chatcodereasoning

No models found

Try adjusting your search or filters

Reset filters

Data sourced from llama.cpp, Ollama and LM Studio.

Built by midudev