# CanIRun.ai

> A web tool that helps you find out which AI models your machine can actually run locally, based on your GPU, VRAM, and memory bandwidth.

CanIRun.ai is a free web tool that lets you instantly check which open-source AI models are compatible with your local hardware. By selecting your GPU or Apple Silicon chip, it calculates VRAM usage, estimated inference speed (tokens/second), and assigns a runability score to hundreds of models. The tool covers models from Meta, Mistral, Google, Alibaba, DeepSeek, and more, with support for multiple quantization formats (Q2_K through F16/GGUF).

- **Hardware compatibility checker**: *Select your GPU (NVIDIA, AMD, Intel, Apple Silicon, Qualcomm, etc.) or set custom VRAM/bandwidth values to see which models fit your machine.*
- **Runability scoring**: *Each model receives a score (0–100) and a grade (Runs great / Runs well / Decent / Tight fit / Barely runs / Too heavy) based on your hardware profile.*
- **Quantization format breakdown**: *View estimated file sizes and quality retention for Q2_K, Q3_K_M, Q4_K_M, Q5_K_M, Q6_K, Q8_0, and F16 formats for every model.*
- **Model filtering and sorting**: *Filter by task (chat, code, reasoning, vision), provider, license, and architecture (Dense/MoE); sort by score, parameter count, context length, speed, or VRAM.*
- **Model comparison**: *Use the compare feature to evaluate multiple models side-by-side across hardware profiles.*
- **Tier list**: *Browse a ranked tier list of models to quickly identify the best options for your hardware class.*
- **Educational docs**: *Built-in documentation explains parameters, quantization, VRAM, MoE architecture, context length, tokens/second, GGUF format, and memory bandwidth.*
- **Data sourced from llama.cpp, Ollama, and LM Studio**: *Model data is kept up to date and reflects real-world local inference tooling.*

## Features
- GPU compatibility checker
- VRAM usage estimation
- Tokens per second estimation
- Runability scoring (0–100)
- Quantization format comparison (Q2_K to F16)
- Model filtering by task, provider, license, architecture
- Model sorting by score, params, context, speed, VRAM
- Model comparison tool
- Tier list view
- MoE and Dense architecture support
- Apple Silicon support
- Educational documentation on AI model concepts
- WebGPU-based hardware detection

## Integrations
llama.cpp, Ollama, LM Studio, GPT4All, HuggingFace

## Platforms
WEB

## Pricing
Free

## Links
- Website: https://www.canirun.ai
- Documentation: https://www.canirun.ai/docs
- Repository: https://github.com/ggerganov/llama.cpp
- EveryDev.ai: https://www.everydev.ai/tools/canirun-ai