General · Open source · Apache 2.0
Kedy Orion
A 1B parameter general-purpose model — balanced, fast, and instruction-tuned for everyday tasks like chat, summarization, and question answering.
What it's built for
Conversational AI
Power chatbots, virtual assistants, and customer support with natural dialogue.
Summarization
Condense long documents, articles, and reports into clear, concise summaries.
Q&A
Answer questions over documents, knowledge bases, and unstructured text.
Content creation
Draft emails, blog posts, product descriptions, and social copy at scale.
Benchmarks
57-subject multiple choice exam spanning STEM, humanities, law, medicine, and more.
Commonsense NLI — predicting the correct continuation of an everyday scenario.
Hard science questions that require knowledge and reasoning beyond pattern matching.
Measures tendency to reproduce common human misconceptions and hallucinations.
Commonsense pronoun resolution requiring world knowledge and reasoning.
Multi-turn conversation quality judged by GPT-4 across 8 capability categories.
Quick start
# Install Kedy CLI pip install kedy # Pull the model kedy pull orion-1b # Run inference from kedy import Model model = Model("orion-1b") response = model.chat([ {"role": "user", "content": "Summarize the key points of the Paris Agreement."} ]) print(response.text)
Ready to use Orion?
Free and open source. Runs on your hardware. No account required to download.