aryabhatta-01-sft is small math reasoning model built on top of Qwen3-1.7B. It was tuned on synthetic AIME-style problems with worked solutions to improve high school olympiad-level math.

Intended use

Solving or checking high school level problems
Assisting with algebra, geometry, or combinatorics..

Example Usage

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("think-a-tron/aryabhatta-01-sft")
tok = AutoTokenizer.from_pretrained("think-a-tron/aryabhatta-01-sft")

prompt = "Problem: What is the sum of all positive divisors of 360?"
inputs = tok(prompt, return_tensors="pt")
out = model.generate(**inputs, max_new_tokens=256, temperature=0.0)
print(tok.decode(out[0], skip_special_tokens=True))

Limitations

Trained only on synthetic data; real contest benchmarks should be kept for evaluation