Back to Tool
X

Xyntetik Runner

Tool Oct 17, 2026

An independent local C engine for GGUF inference, LoRA adaptation and reproducible run records.

Visit primary link 2 sources attached
Xyntetik Runner's diagram contrasts training a separate 16-bit model with training an adapter through the same 4-bit file used for serving.
Author-published workflow diagram from the tagged 1.4.2 English guide. It illustrates the project's approach, not an independent reproducibility test. Full frame preserved; the release limits bit-for-bit rollout identity to CPU sampling with one-token prompt processing. Source: Xyntetik Runner / Joakimpalm-Zen
agent-ecosystemself-hostedai-agents

What happened lately

Xyntetik Runner timeline

Xyntetik Runner serves supported GGUF files locally and trains LoRA adapters through the quantized weights it uses for inference. The Apache-2.0-licensed project provides binaries for macOS Apple Silicon, Linux and Windows, with an OpenAI-compatible local API among its documented endpoints. It is separate from the coding-agent clients that can connect to it.

Version 1.4.2 changes the reinforcement loop to preserve sampled token IDs and records the starting adapter for continued training. Its bit-for-bit rollout claim applies to CPU sampling with one-token prompt processing; GPU sampling and CPU-hosted training have different limits. Capabilities and test outcomes are author-reported.