Neural weights
FP16
Safetensors release for compatible Transformers workflows.
Explore model files ↗Official model showcase · Version 1
Meet Alphi v1.
A public Gemma 4 E2B-it-based AI model release from Ryzova. Explore full-precision weights, GGUF variants, research notes, and reproducibility resources.
Choose your model format
Download verified release artifacts from the model repository. GPU or CPU suitability depends on your available memory and the inference engine's Gemma 4 support.
GGUF
F16
Higher-memory GGUF variant, suited to precision-focused testing.
Browse F16 ↗GGUF
Q4_K_M
Compact quantized release for more accessible local use.
Browse Q4 ↗GGUF
Q6_K / Q8_0
Alternative quality and memory trade-offs for local inference.
Browse variants ↗Evaluation status
Transparent reporting matters more than inflated numbers.
● Compatibility diagnostic completed
F16 GGUF inference and LM-Evaluation Harness compatibility checks completed successfully using a limited diagnostic run. The test used ARC-Easy and IFEval samples.
Inference: checkedHarness: checkedFull scores: pending
Not a benchmark score claim
A small-sample diagnostic is not statistically meaningful evidence of full task performance. Full standardized benchmark scores are not yet established. Comparative model performance claims are intentionally omitted.
Neural model and broader Ryzova research
Ryzova explores Matrix Memory, Growing Brain, retrieval and agent tools as broader system-level components. These should not be confused with capabilities automatically bundled into the standalone model weights.
Model ID:
RyzovaTech/Alphi-v1 Looking for a live chat demo?
This Space is an official information hub, not a live model inference service. It does not generate or simulate Alphi responses. Serving Alphi interactively requires a compatible inference backend and compute resources.