Profile Avatar

Support Open Source

Support Our Open-Source Releases - models, datasets, and more!

☕ Support on Ko-fi

🚀 New Releases

Esper 4

Model

Agentic coding, architecture, DevOps, and MLOps finetune for Qwen 3.6, Gemma 4, and Muse Glimmer!

Tachibana-Agent

Model

Agentic coding finetune of Qwen 3.6, Gemma 4, and Muse Glimmer using the new Tachibana 4 dataset.

Mitakihara 2

Dataset

AI development, research, deployment, interpretability, operation and experimentation dataset.

Titanium 4

Dataset

Agentic DevOps and architecture dataset generated with DeepSeek-V4-Pro.

Tachibana 4

Dataset

Agentic coding dataset generated with DeepSeek-V4-Pro.

Esper 3.1

Model

Esper 3.1 now available for Qwen3.6-35B-A3B and Qwen3.6-27B, plus other small models!

📚 Collections

Experimental Reasoning Models

Collection

Experimental fine-tuned models that utilize reasoning to create specific structured forms of output.

Reasoning Datasets

Collection

Synthetic datasets generated using reasoning models, primarily the DeepSeek-V4, Deepseek-R1 and Deepseek-V3 series.

🛠️ Coming Soon

  • ⏳
    Micromodels: Base and Chat Micro-sized models, trained from scratch on high quality tokens.
  • ⏳
    Micromodels: Experimental Releases The fun part :) we're going to get creative with micro releases that we can make look very different from mainstream LLMs!
  • ⏳
    Esper 4 Qwen 3.8 and more models to come for Esper 4: donate to see more releases!
  • ⏳
    Celestia 4 Science, agentic. exciting!
  • ⏳
    Shining Valiant 4 A new agentic approach. We're very excited about this one!
  • ⏳
    Experimental Token Compression Releases Multiple experimental releases on reasoning compression and structure.