OpenAI releases o3, extending the reasoning paradigm. Frontier math and coding benchmarks (ARC-AGI, FrontierMath) show dramatic improvement.

Anthropic releases Claude 3.7 Sonnet with explicit thinking mode. Users can see the models reasoning trace, setting a new transparency bar for frontier models.

Meta announces Llama 4, its first open-source natively multi-modal reasoning model. Open-source AI continues to close the gap with frontier closed models.