Always ship performant code

Codeflash uses AI to automatically find the most performant version of your Python code through benchmarking—while verifying it's correct

pip install codeflash && codeflash init
pip install codeflash && codeflash init
Trusted by Top Engineering Teams at
AirBytepydanticlangflow logo

Optimize most Python code you write

Codeflash generates several ideas for optimization and finds the best one that is the fastest and correct.
Codeflash solves the problem, not just create "ai suggestions".

card1a
298x Speedup for Langflow by eliminating a loop, and deduplicating comparisons. Link to PR
34% Speedup for Pydantic by reducing recursion overhead by managing an explicit stack. Link to PR
89x Speedup by reducing algorithmic complexity by converting a list implementation to set.
148% Speedup for LangChain by not sorting a large array and instead using argpartition. Link to PR

We've used Codeflash in the Pydantic codebase (300M+ downloads / month) to help optimize logic including repeated attribute access patterns and recursive algorithms. We're particularly impressed by the thorough testing and explanations provided by Codeflash which give us confidence in merging the recommended changes.

Sydney Runkle
Sydney Runkle

Software Engineer @ Pydantic

Keeping a large open-source project performant can be quite hard to do. With Codeflash we can be more confident that every new PR is optimized.

Gabriel Almeida
Gabriel Almeida

Founder & CTO @ Langflow

I've been working on Albumentations, an image augmentation library, for years now. The library is already well optimized. But the Codeflash team submitted over 10 pull requests in just one week, each making our functions significantly faster. The speedups range from 2× to 55× in some cases. I'm now adding Codeflash to our GitHub actions workflow so every PR will automatically get checked for potential performance improvements. It will help make the library even faster and teach everyone who contributes to Albumentations how to write more performant code.

Vladimir Iglovikov
Vladimir Iglovikov

Founder and CEO @ Albumentations.ai

We've used Codeflash in the Pydantic codebase (300M+ downloads / month) to help optimize logic including repeated attribute access patterns and recursive algorithms. We're particularly impressed by the thorough testing and explanations provided by Codeflash which give us confidence in merging the recommended changes.

Sydney Runkle
Sydney Runkle

Software Engineer @ Pydantic

Keeping a large open-source project performant can be quite hard to do. With Codeflash we can be more confident that every new PR is optimized.

Gabriel Almeida
Gabriel Almeida

Founder & CTO @ Langflow

I've been working on Albumentations, an image augmentation library, for years now. The library is already well optimized. But the Codeflash team submitted over 10 pull requests in just one week, each making our functions significantly faster. The speedups range from 2× to 55× in some cases. I'm now adding Codeflash to our GitHub actions workflow so every PR will automatically get checked for potential performance improvements. It will help make the library even faster and teach everyone who contributes to Albumentations how to write more performant code.

Vladimir Iglovikov
Vladimir Iglovikov

Founder and CEO @ Albumentations.ai

Write Highly Performant Code - While Shipping Features Fast

card5

Finding and optimizing slow code is time consuming in sprawling code bases

card6

Codeflash helps developers spend more time doing what they love: shipping

card7

Continuously Optimize your GitHub PRs

Install Codeflash as a GitHub action and never ship slow code again.

card2a

Optimize entire projects

Run codeflash --all and codeflash acts as an expert performance engineer to profile, optimize and test every code path in your project to make it all blazing fast.

card3a

Find Optimizations in 5 minutes

Discover how much faster your code can be by scanning it with Codeflash.

pip install codeflash && codeflash init
pip install codeflash && codeflash init
cta graphic base