Can I run AI locally?
canirun.ai
March 13, 2026
1 min read
🔥🔥🔥🔥🔥
79/100
Summary
Detect your hardware and find out which AI models you can run locally. GPU, CPU, and RAM analysis in your browser.
Community Sentiment
Positives
- MoE models like GPT-OSS-20B can produce more tokens per second on the same hardware, making them efficient for local AI applications.
- Recent models like Qwen 3.5 show significant improvements in performance, indicating that advancements in AI are leading to better results even with smaller models.
Concerns
- The estimation method for model performance based on memory bandwidth may not accurately reflect the capabilities of MoE models, leading to potential misunderstandings.
- Perplexity is a more critical metric than tokens per second, as high-speed outputs can still yield poor quality results, emphasizing the importance of quality over speed.