According to @DeepLearningAI, Anthropic launched Claude Sonnet 4.5 with a variable reasoning-token budget, expanded 200k–1M token context windows, improved coding and reasoning benchmark performance, ...