Scientific · Open source · Apache 2.0
Kedy Maxwell
A 1B parameter model trained on scientific literature, mathematical reasoning, and STEM domains — compact enough to run locally, precise enough to trust.
What it's built for
Research assistance
Summarize papers, extract methods, and synthesize findings across literature.
Math & equations
Step-by-step solutions for algebra, calculus, statistics, and formal proofs.
Data analysis
Interpret results, suggest statistical methods, and explain experimental design.
STEM education
Explain complex concepts, generate practice problems, and tutor students.
Benchmarks
Competition math problems requiring multi-step reasoning across 7 difficulty levels.
Grade school math word problems requiring 2–8 step arithmetic reasoning.
Graduate-level expert questions in biology, chemistry, and physics.
STEM subset of Massive Multitask Language Understanding covering 20 science domains.
Science questions from natural science exams and educational resources.
Hard science questions requiring knowledge beyond simple retrieval or co-occurrence.
Quick start
# Install Kedy CLI pip install kedy # Pull the model kedy pull maxwell-1b # Run inference from kedy import Model model = Model("maxwell-1b") response = model.generate( "Derive the quadratic formula from ax² + bx + c = 0.", max_tokens=512 ) print(response.text)
Ready to use Maxwell?
Free and open source. Runs on your hardware. No account required to download.