Show HN: AQ – a from-scratch 1B academic LLM by a 2-person team in India

zyoraclub2 pts0 comments

zyoralabs/AQ-academic-ai · Hugging Face

Log In<br>Sign Up

","eos_token":"","pad_token":""}},"createdAt":"2026-07-25T04:39:40.000Z","discussionsDisabled":false,"discussionsSorting":"recently-created","downloads":12,"downloadsAllTime":12,"id":"zyoralabs/AQ-academic-ai","isLikedByUser":false,"availableInferenceProviders":[],"showHuggingChatEntry":false,"inference":"","lastModified":"2026-08-03T08:34:34.000Z","likes":0,"pipeline_tag":"text-generation","librariesOther":[],"trackDownloads":true,"model-index":null,"private":false,"repoType":"model","gated":false,"tags":["safetensors","aq","education","academic","tutor","concept-first","india","from-scratch","text-generation","custom_code","en","ta","hi","license:apache-2.0","region:us"],"tag_objs":[{"id":"text-generation","label":"Text Generation","type":"pipeline_tag","subType":"nlp"},{"id":"safetensors","label":"Safetensors","type":"library"},{"id":"en","label":"English","type":"language"},{"id":"ta","label":"Tamil","type":"language"},{"id":"hi","label":"Hindi","type":"language"},{"id":"aq","label":"aq","type":"other","clickable":true},{"id":"education","label":"education","type":"other","clickable":true},{"id":"academic","label":"academic","type":"other","clickable":true},{"id":"tutor","label":"tutor","type":"other","clickable":true},{"id":"concept-first","label":"concept-first","type":"other","clickable":true},{"id":"india","label":"india","type":"other","clickable":true},{"id":"from-scratch","label":"from-scratch","type":"other","clickable":true},{"id":"custom_code","label":"custom_code","type":"other","clickable":true},{"id":"license:apache-2.0","label":"apache-2.0","type":"license"},{"type":"region","label":"🇺🇸 Region: US","id":"region:us"}],"widgetData":[{"text":"My name is Julien and I like to"},{"text":"I like traveling by train because"},{"text":"Paris is an amazing place to visit,"},{"text":"Once upon a time,"}],"safetensors":{"parameters":{"BF16":1257260544},"total":1257260544,"sharded":false,"totalFileSize":2514566480},"hasBlockedOids":false,"region":"us","isQuantized":false},"discussionsStats":{"closed":0,"open":0,"total":0},"query":{},"inferenceContextData":{"billableEntities":[],"entityName2Providers":{}},"hasQuantizations":false,"copyToBucketNamespaces":[]}">

AQ-Academic-AI — Academic Quotient v1 (Tutor)

AQ (Academic Quotient) — India's Concept-First Academic AI. v1 now live.<br>Raising the Academic Quotient of every student.

AQ is a 1.26B-parameter academic tutor built completely from scratch by Zyora Labs — proprietary architecture, own training code (pure PyTorch), own tokenizer, own data pipeline, own tutor fine-tune. No fine-tune of any existing model.

This is the tutor (instruct) model : it answers student questions directly with explanations, numbered steps, and worked examples. The pretrained base model is available at zyoralabs/AQ-academic-ai-base.

Training

Pretraining — 20B tokens , knowledge-dense and concept-first: real textbooks, course notes, scientific papers, encyclopedic text, mathematical reasoning — in English, Tamil, and Hindi. Grown progressively 75M → 300M → 1.26B, finished with a quality anneal (LR → 0 on the highest-quality academic text).

Tutor fine-tune — 400M tokens of educator-style instruction data (explanations, step-by-step math, knowledge Q&A, Hindi instructions), loss masked to tutor responses.

Prompt format

### Student:<br>{your question}

### Tutor:

from transformers import AutoModelForCausalLM, AutoTokenizer

tok = AutoTokenizer.from_pretrained("zyoralabs/AQ-academic-ai")<br>model = AutoModelForCausalLM.from_pretrained("zyoralabs/AQ-academic-ai", trust_remote_code=True)

prompt = "### Student:\nWhat is a stack in data structures?\n\n### Tutor:\n"<br>ids = tok(prompt, return_tensors="pt").input_ids<br>out = model.generate(ids, max_new_tokens=200, do_sample=True,<br>temperature=0.7, top_k=40, repetition_penalty=1.3)<br>print(tok.decode(out[0][ids.shape[1]:]))

Architecture (proprietary, from scratch)

Parameters<br>1.26B

Layers<br>48

Hidden size<br>1536

Attention heads<br>24 (grouped-query, 8 KV heads)

Feed-forward<br>SwiGLU, 4096

Positional encoding<br>Rotary (RoPE)

Normalization<br>RMSNorm

Context length<br>2048

Vocabulary<br>32,000 (byte-level BPE, English + Tamil + Hindi)

Benchmarks (0-shot, lm-evaluation-harness)

Benchmark<br>AQ v1 Tutor<br>Notes

SciQ<br>69.2<br>strong science knowledge for the size/data budget

PIQA<br>62.3

ARC-easy<br>45.1

Winogrande<br>49.5

HellaSwag<br>29.6

MMLU<br>25.2<br>at-chance, like all ~1B-class models

ARC-challenge<br>21.2

For context: models of this size trained on 15×–150× more tokens (e.g. 300B–3T) reach SciQ ~84–89. AQ reaches ~70 on just 20B tokens — the concept-first, knowledge-dense corpus is the point.

Transparency: as a 1.26B model, AQ v1 has real limits — multi-step arithmetic word problems and MMLU-style abstract reasoning are weak (these unlock at larger scale, on our roadmap). In the AQ product, answers are additionally grounded with retrieval over real study material.

Intended use

The tutor layer of the AQ...

academic label type tutor true text

Related Articles