# Maxwell

> Maxwell improves code while it runs, adapting to hardware and data. Every improvement is formally proven equivalent to the original code.

Maxwell is the first Code Execution Model (CEM): it lives inside a running Python program, replaces code with faster hardware-optimized code, and accompanies every replacement with a machine-checked proof of semantic equivalence. It is made by [doubleAI](https://www.doubleai.com).

Benchmarks:

- 38x faster data processing pipelines
- 4.25x faster AI inference kernels

## Video

- [What is a code execution model?](https://heymaxwell.ai/video/what-is-a-code-execution-model.mp4): A short animated film, with narration and captions, introducing Maxwell: most code never sees the hardware it runs on or the data it processes. Maxwell watches a program while it runs, rewrites the binary underneath to fit that exact machine and data, and formally proves every improvement equivalent to the original. It runs on CPUs and GPUs, speeding up data pipelines and AI inference kernels.

## Demos

- [GPU-based speech-to-text using Whisper](https://heymaxwell.ai/demo/whisper-gpu.mp4): Two terminals run the same Python script on the same RTX 4090: 180 transcriptions with Whisper-small from a cold start. Without Maxwell, on the left, it takes 106.23 seconds. With Maxwell, on the right, it completes in 49.82 seconds. Both produce identical transcripts. The recording plays at 4x speed.
- [CPU-based graph algorithms](https://heymaxwell.ai/demo/graph-cpu.mp4): Two terminals run the same Python script on the same machine: NetworkX betweenness centrality, 15 runs. Without Maxwell, on the left, it takes 61.44 seconds. With Maxwell, on the right, it completes in 17.50 seconds. Both print identical results. The recording plays at 4x speed.

## Research

- [Maxwell approaches Speed of Light on Blackwell](https://www.doubleai.com/research/maxwell-approaches-speed-of-light-on-blackwell): After one day of search, Maxwell beats NVIDIA's PyTorch baselines on 90% of SOL-ExecBench's Blackwell kernels.
- [doubleAI's Maxwell: Surpassing Expert-Written Kernels At Scale](https://www.doubleai.com/research/doubleais-maxwell-surpassing-expert-written-kernels-at-scale): Independently generated and verified a hyper-optimized drop-in version of NVIDIA cuGraph, 3.6x faster on average.
- [Blazing Fast pandas Without Changing a Line](https://www.doubleai.com/research/maxwell-pandas-blogpost): Maxwell completely rewrote pandas' kernels from scratch, specialized for modern hardware.

## Manifesto

- [All Thinking Things](https://heymaxwell.ai/manifesto): the manifesto behind Maxwell

## Case study

- [LTX-2.5 at 4K, in half the time](https://heymaxwell.ai/customers/ltx): LTX-2.5 is LTX's state-of-the-art open world model, which generates, simulates, and shapes video, audio, and the physical world. Maxwell made LTX-2.5 generate 4K video nearly twice as fast, with the same output.

## Optional

- [Get early access](https://heymaxwell.ai/early-access)
- [Privacy Policy](https://heymaxwell.ai/privacy-policy)
- [Terms of Use](https://heymaxwell.ai/terms)
- [Contact](https://heymaxwell.ai/contact)