Introducing AXIOM 2: Our Most Capable and Safest Model
Today we're releasing AXIOM 2, a significant advance in both capability and safety. This model achieves new state-of-the-art results across scientific reasoning, code, and mathematics while passing our most rigorous safety evaluations.
Today marks the release of AXIOM 2, our second-generation frontier model. This release represents over eighteen months of research across capabilities, safety, and interpretability.
Capability Improvements
AXIOM 2 achieves state-of-the-art performance on:
- Scientific reasoning: 89.4% on GPQA Diamond
- Mathematics: 94.2% on MATH-500
- Code: 72.1% on SWE-bench
- Multilingual: Supports 52 languages at near-native quality
Safety Advances
Every capability gain is matched by rigorous safety work. AXIOM 2 passes all current frontier safety evaluations and introduces new capabilities for expressing uncertainty and refusing harmful requests while remaining maximally helpful.
API Availability
AXIOM 2 is available today via our API. See our documentation for integration guides and pricing.