What’s happening in AI today.

The flagship model now handles 1M tokens, adds structured output improvements, and reduces tool-call latency by ~28%. Priced at parity with 4.6. Early benchmarks show meaningful gains on long-horizon agent tasks.