We’re preparing to release AstroSage-8B (de Haan et al., in prep.), our flagship 8B parameter model for astronomy.

• Top-performing 8B model in astronomy Q&A
• Rivals 70B models in astronomy Q&A compared to Qwen-2.5-72B (and GPT-4o).
• Aims for affordability and accessibility in astronomy research

We’re currently:

1. Quantizing the model for memory efficiency
2. Optimizing for Ollama to run on standard laptops
3. Benchmarking performance on local machines

Check back in a month!