magnitudedev/magnitude

This page introduces an external open-source repository. It is not an HDATF product.

What it does

An inference engine for agents that compiles and tunes kernels on the user's hardware. It runs open models on Apple Silicon, NVIDIA, AMD or CPU and connects to existing agents through a desktop app and CLI.

How it helps ATF

Harness could compare its local inference and agent connection approach as an execution backend candidate. Hardware-specific tuning and shared prefix caches are useful comparison points.

License

Apache-2.0 Permissive, with a patent grant. Commercial use is allowed; keep the notices and state your changes.

More in this category

• ProjectDMX/DMI
  A decoupled, asynchronous observability backend for LLM inference and training that exposes internal model states. It is a research preview with HuggingFace, vLLM and Megatron-LM integrations, and its APIs may change.
• ninjahawk/livenerf
  A longitudinal benchmark for detecting model capability drift after release. It uses frozen prompts, a pinned CLI, fixed graders and retained raw logs.
• deepopen-com/deepopen
  A non-autoregressive engine built on Laya for multilingual structured decisions. It performs multidimensional classification in a single forward pass.
• basetenlabs/baseten-switch
  A local macOS app and gateway routing supported coding harnesses between native providers and Baseten models. It has a global routing switch and per-client model mappings and remains in beta.
• firelex/jeff
  Fine-tuned Qwen3.5 and Gemma 4 models classify situations against options described in plain language and return probabilities without generating text.

Only repositories in the ranked Trendshift lists are included, and the lists are used only to find candidates. We do not copy their ranks. Descriptions, licenses and star counts come from each GitHub repository. The notes are our own reading. We have not tested these projects, and a place on a trending list does not prove quality.