​
Use cases

Thunder Compute is optimized for AI/ML development workflows. That said, Thunder Compute has the full functionality of an EC2-style on-demand GPU cloud instance.

​
Supported Software

The following software is tested and supported on Thunder Compute. For a list of pre-installed libraries and their exact versions, see our Instance Details guide.

​
Core Environment

  • CUDA: We support version 12.0 or greater.
  • CUDNN: We support version 9.0 or greater.

Do not attempt to reinstall CUDA. If it seems like you need an older CUDA driver, you are almost always better off upgrading your other dependencies (e.g., PyTorch).

​
AI/ML Libraries

  • PyTorch: We include version 2.7.1, downgrading can cause issues.
  • Notebooks: JupyterLab and Jupyter Notebooks are supported.
  • Other Tools: AI model serving tools like ComfyUI, Ollama, VLLM, and Unsloth are well-tested.

​
Experimental (less stable)

The following workloads are less tested, experimental, or unstable:

  • Tensorflow [experimental]
  • PyTorch Lightning [experimental]
  • Jax [experimental]
  • Custom CUDA Kernels [unpredictable behavior, particularly with errors and profiling. Message us for details]

If you encounter compatibility issues with these experimental workloads, consider switching to Production mode for maximum compatibility and predictable performance.

​
Unsupported

Currently, Thunder Compute lacks official support for graphics workloads such as OpenGL, Vulkan, and FFMPEG. If you’d like to run these, contact us.

For workloads requiring graphics support, custom kernels, or components incompatible with the prototyping tier, use Production mode which provides maximum stability and reliability with all optimizations disabled.

​
Cryptocurrency mining

Mining, staking, or otherwise interacting with cryptocurrency is strictly prohibited on Thunder Compute. If cryptocurrency-related activity is detected, the associated account is immediately banned from Thunder Compute and any billing credit is revoked. The account is then billed for the full amount of usage.

​
Benchmarking

If you are benchmarking (temperature, wattage, etc.), do not use prototyping instances. Due to the optimizations we use, temperature, wattage, and utilization numbers are not representative. Comparing model performance (iterations / second) is the best apples-to-apples metric.

​
Geographic availability

Thunder Compute is only available for B2B customers, i.e., requires a VAT ID (or similar) in the following countries:

  • United Arab Emirates
  • Angola
  • Bahrain
  • Brazil
  • Switzerland
  • Côte d’Ivoire (Ivory Coast)
  • Colombia
  • Algeria
  • Georgia
  • Iraq
  • Jordan
  • Kazakhstan
  • South Korea (Republic of Korea)
  • Kuwait
  • Morocco
  • North Macedonia
  • Oman
  • Paraguay
  • Qatar
  • Saudi Arabia
  • Tunisia
  • Turkey (Türkiye)
  • Tanzania
  • Ukraine
  • Uganda
  • Uzbekistan
  • Yemen
  • India
  • Moldova (Republic of Moldova)

Thunder Compute is not currently available in the following countries:

  • Belarus
  • China
  • Cuba
  • Indonesia
  • Iran
  • Kenya
  • North Korea
  • Malaysia
  • Mexico
  • Nigeria
  • Russia
  • Sudan
  • Syria
  • Uruguay

If you’re located in one of these countries and need access to Thunder Compute, please contact us to discuss potential alternatives.

​
Miscellaneous tips

We use a new kind of virtualization to maximize GPU utilization, reducing your cost. To learn more about how this works, check out this blog post.

If you encounter any strange issues or errors, please check our troubleshooting guide or contact us.

To help you get started with Thunder Compute, we recommend checking out these guides: