Anthropic Releases Claude 3.7: Best Coding and Scientific Reasoning Model

Anthropic released Claude 3.7, achieving new state-of-the-art on coding benchmarks (HumanEval: 95.2%, SWE-Bench: 49.5%) and scientific reasoning. The model features an improved "extended thinking" mode for complex multi-step problems, 200K context maintained with improved mid-context accuracy, and new computer use capabilities for browser automation. Anthropic's model card transparency reports show Claude 3.7 has the lowest hallucination rate in independent evaluations. Available on Claude.ai and API at same pricing as Claude 3.5.