nace-ai/kapitsa v0.1.37 is a release that refreshes the repository introduction of an inference engine optimized for Qwen and Nemotron on Apple Silicon. According to the GitHub release notes, the repository title, intro image, and command-line screenshot were changed to match the engine description, and a benchmark for a MacBook Pro M1 Pro 16 GB environment was added. A cautionary note that had been displayed while other apps were using the graphics processing unit was removed from the README.
The project is presented as an Apple Silicon-optimized inference engine that runs Qwen and Nemotron models. It advertises OpenAI and Anthropic API compatibility, making it an option for connecting models in tools that use those APIs. The focus of this change is less on new inference capabilities than on aligning the repository's description and visuals with the project introduction, and on adding benchmark information for a specific MacBook Pro environment.
The provided release information gives no benchmark result numbers or measurement conditions. So this addition alone does not let you judge speed differences against other devices or models. What can be confirmed in this release is the removal of the README caution, the tidy-up of the intro materials, and the addition of the benchmark for that environment. It is closer to a version that updates introduction and documentation information than one that announces a performance improvement.