antirez/ds4

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

What it does

DwarfStar, a small native inference engine for running a few large language models such as DeepSeek V4 Flash and GLM 5.x on consumer hardware, targeting Metal, CUDA and ROCm. It uses only the GGUF files the project produces.

How it helps ATF

Worth comparing for AX consulting cases where a model has to run on local hardware. Because it builds and tests tool calls, the HTTP server and its coding agent together, it could also be a reference for a local executor under Harness.

License

MIT Permissive. Commercial use and changes are allowed if the copyright notice is kept.

More in this category

• tashfeenahmed/freellmapi
  FreeLLMAPI puts free tiers from many LLM providers behind one OpenAI-compatible /v1 API. A router picks an available model, switches providers when one is rate-limited, and tracks usage per key to stay under free-tier limits.
• decolua/9router
  9Router is a router that connects AI coding tools such as Claude Code, Codex, Cursor and Cline to more than 40 providers. It falls back from subscription to cheap to free models and compresses tool results to save tokens.
• google-research/timesfm
  TimesFM is a pretrained foundation model from Google Research for time-series forecasting. The latest version, 3.0, adds multivariate forecasting and support for past and future covariates.
• kyegomez/OpenMythos
  A community reconstruction, based on public research and speculation, of the Claude Mythos model architecture, not affiliated with Anthropic. It implements a recurrent-depth transformer with a looped block, MLA or GQA attention and mixture-of-experts layers.
• diegosouzapw/OmniRoute
  OmniRoute is an MIT-licensed AI gateway that puts many model providers behind one endpoint for tools such as Claude Code, Codex, Cursor and Cline. It offers quota-aware fallback, token compression and a dashboard tracking provider free tiers.

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.