RaspizdAI

Welcome to RaspizdAI. We are an open-source research collective focused on training compact language models from scratch, creating niche datasets, and building custom evaluation benchmarks. We move fast, break conventions, and explore unconventional AI concepts just for the fun of it.

No corporate fluff, no over-engineering—just raw models and experiments built to see what happens.

What We Do

• Models From Scratch: Compact, lightweight LLMs designed, structured, and trained entirely from the ground up.
• Fine-Tunes: Unfiltered, specialized model variants pushed past standard guardrails and default behaviors.
• Datasets & Benchmarks: Niche, raw datasets paired with custom evaluation benchmarks to test model limits in unconventional scenarios.

Disclaimer: Expect experimental approaches, edge-case evaluations, and zero polished corporate assurances. Everything here is provided as-is—use at your own risk.

Avatar made from assets sourced from drednot.io (not affiliated; rights to owners; will remove upon request).