Nova is a family of compact, open-weight language models from HyperAiCorp, designed for efficient local inference and robust tool use.
AI & ML interests
We are working on fine-tuning LLM, creating our own datasets and possibly benchmarks⚡.
Recent Activity
Organization Card
HyperAI
Advancing open and reproducible AI through fine-tuning, datasets, and benchmarks.
HyperAI is an AI research organization focused on fine-tuning large language models, building high-quality custom datasets, and developing transparent benchmarks.
Focus areas
- LLM fine-tuning — adapting open foundation models to specialized tasks and domains
- Custom datasets — curating and building high-quality training data
- Benchmarks — developing reproducible evaluation suites to measure model performance
Flagship model
- Nova-1.1-0.8B — a fine-tuned conversational assistant built on Qwen3.5-0.8B with protocol-agnostic function calling (XML, JSON, plain-text tools), multi-turn tool chains, and preserved general knowledge across math and science benchmarks.
Get involved
- Follow our models and datasets to catch the latest releases
- Open an issue or discussion to collaborate with us
Contact
- Organization: HyperAiCorp
datasets 0
None public yet