Benchsift
- 5 Devlogs
- 10 Total hours
My website allows comparing several AI models using multiple benchmarks. The models are retrieved via the OpenRouter and Artificial Analysis APIs, as well as the benchmarks.
Hey everyone!First off, thanks so much for all the feedback! I’ve made a few updates to the site based on your suggestions:Sorting & Categorization: Separated the Open Weight filter from the Intelligence filter (they used to be in the same spot).Header: Made the header more compact.Navigation: Moved DeepSWE and Coding Agent under the Other Benchmarks section.New Page: Added an “About” page explaining what BenchSift is and providing a bit more context on the project.
Ship #1
update page 404
Improved the comparison page when no model is selected, as well as the catalog. Added several new benchmarks and fixed a bug with theme switching. Updated the model pages styling to make them clearer and more readable, and adjusted the benchmark colors. Models are now grouped together previously they were split by reasoning level.
New icon and website optimization
Added two modes: a normal mode which is a simple ranking mode based on category, and a nerd mode which is more comprehensive with more data. Fixed the desktop and mobile UI. UX has also been improved, as well as site speed. It’s also lighter now.
