You can deploy this model and start earning money today!

zyoralabs/
AQ-academic-ai-GGUF

Name: AQ-academic-ai-GGUF
Availability: InStock

zyoralabs/AQ-academic-ai

quantized

apache-2.0

endpoints_compatible

Views

No views yet

AQ Academic AI GGUF — English Edge Preview

This repository contains an experimental English-only edge preview of zyoralabs/AQ-academic-ai, a 1.26B-parameter academic tutor built from scratch by Zyora Labs.

Artifact

File	Quantization	Size	Context
`aq-tutor-v2-q4_k_m.gguf`	Q4_K_M	765,358,592 bytes (~730 MiB)	2,048 tokens

SHA-256:

b675e036ace4934835bf323ee43fc39dca0d2fbb5d3c0435e3c680c2ea4aa278

Usage

bash

llama-cli \
  -m aq-tutor-v2-q4_k_m.gguf \
  -p $'### Student:\nWhat is photosynthesis?\n\n### Tutor:\n' \
  -n 200 --temp 0.7 --top-k 40 --repeat-penalty 1.3 --ctx-size 2048

Validation

All 434 source tensors (1,257,260,544 values) matched the canonical package exactly.
BF16 AQ and canonical Llama forward logits matched exactly on English, Tamil, and Hindi prompts.
Q4 English smoke checks passed for mathematics and science.
Example results: derivative of x² → 2x; photosynthesis → coherent academic explanation.
Source model revision: 19482cacfa3a1157fce7a51a22285ec10f689133.
llama.cpp revision: 153d324bcf86d220b235ca010eeb11213f32b5d1.

Important limitation

This Q4 artifact is currently released as an English-only preview. Tamil and Hindi generation did not pass release-quality checks. Do not represent this GGUF as a validated multilingual release. The authoritative BF16 model remains in the source repository.

Quantization creates approximate derivative weights; it does not alter the original AQ weights.

zyoralabs/AQ-academic-ai-GGUF