Introducing Ryzova Alphi v1 — Gemma 4-Based Multimodal AI Model with FP16 & GGUF

Ryzova released Ryzova Alphi v1, a multimodal AI model built on Google Gemma 4 E2B-it, publishing FP16 Safetensors weights and F16, Q4_K_M, Q6_K and Q8_0 GGUF variants on Hugging Face with a research archive DOI. Ryzova said initial inference and compatibility diagnostics via llama.cpp, the LM-Evaluation Harness and ARC-Easy and IFEval samples completed successfully, but full standardized benchmark scores are not yet available and the company is making no comparative performance claims. The model targets English and Sinhala, and Ryzova said its Matrix Memory, Growing Brain, knowledge retrieval and agent tools components are broader research work not verified or packaged within the downloadable GGUF weights.

Hello Hugging Face community We are introducing Ryzova Alphi v1 , a publicly available AI model release from Ryzova, developed using Google Gemma 4 E2B-it as its foundation. Alphi is part of Ryzova’s broader research into local-first AI, structured memory, adaptive intelligence, and agent-based systems. The Hugging Face repository includes: FP16 Safetensors model weights F16 GGUF for full-precision local inference Q4 K M, Q6 K, and Q8 0 GGUF variants for different hardware requirements Gemma 4 multimodal architecture English and Sinhala as target languages Model repository: RyzovaTech/Alphi-v1 · Hugging Face https://huggingface.co/RyzovaTech/Alphi-v1 Research archive DOI: RyzovaTech/Alphi-v1 · Hugging Face https://doi.org/10.57967/hf/11373 Beyond the neural model release, Ryzova is investigating several system-level capabilities: Matrix Memory: Structured and persistent memory systems Growing Brain: Reusable knowledge and capability development Knowledge retrieval: Retrieval-augmented information access Agent tools: Controlled tool execution and multi-step workflows These are broader Ryzova runtime and research components. They should not be interpreted as capabilities independently verified or fully packaged within the downloadable GGUF weights. We have completed initial inference and compatibility diagnostics using: llama.cpp F16 GGUF inference LM-Evaluation Harness ARC-Easy and IFEval diagnostic samples The initial diagnostic completed successfully. However, full standardized benchmark scores are not yet available , and we are not making comparative performance claims at this stage. We would appreciate feedback on: Local inference compatibility Gemma 4 model evaluation Quantized inference behavior Reproducible benchmarking English and Sinhala testing Potential integration with memory-augmented agents We welcome technical discussion, reproducible test results, and constructive feedback. Thank you for exploring Alphi