Anthropic just dropped Claude 4.6 Opus with a 500K context window, native tool use, and significantly improved reasoning on math and code benchmarks. Early benchmarks show it outperforming GPT-4o on H