Small Doge

SmallDoge is an open-source project that develops compact, high-performance small language models (SLMs) optimized for speed and efficiency. Built with dynamic algorithms, these models are designed for rapid inference while maintaining strong performance—ideal for developers, researchers, and hobbyists who want to run AI locally on modest hardware. The entire training pipeline, code, and model checkpoints are publicly available on GitHub and Hugging Face. The project offers a suite of models: pre-trained bases, supervised fine-tuned (SFT) versions, and reinforcement learning (RL) enhanced variants. It also provides curated multi-stage datasets specifically engineered for SLM training, plus checkpoints for continued training to reduce instability. A downstream applications collection showcases models optimized for real-world tasks. SmallDoge is community-driven and transparent, aiming to democratize AI by removing barriers to entry. The project encourages collaboration via Discord and GitHub. All models are accessible through Hugging Face, making them easy to download and fine-tune on custom data. Unlike larger open-source models, SmallDoge specifically targets resource-constrained environments, trading some capacity for speed and accessibility. It is not a commercial product—there is no API, no SLA, and no enterprise support. For those needing a plug-and-play solution, proprietary offerings like OpenAI or Anthropic are better suited.