HauhauCS Gemma 4 E2B Goes Uncensored
HauhauCS’s 2B-parameter Gemma 4 variant pairs multimodal text, image, video, and audio input with GGUF quantization and reduced safety refusals. It supports local runtimes including llama.cpp, LM Studio, Jan, and Ollama.
This is a compelling local experimentation model, but “uncensored” should not be confused with more capable—the tradeoff is fewer guardrails and limited independent evaluation.
- –Native multimodality brings vision and audio capabilities to a compact 2B model
- –131K context supports long documents, image-heavy prompts, and extended conversations
- –K_P quantizations aim to preserve quality while reducing local hardware requirements
- –GGUF compatibility makes deployment straightforward across popular local inference tools
- –Developers should validate outputs carefully before using it in safety-sensitive workflows
DISCOVERED
2h ago
2026-08-19
PUBLISHED
2h ago
2026-08-19
RELEVANCE
AUTHOR
HuggingModels
