OpenAI: GPT-6 Astra Launches 1M Token Context

OpenAI released GPT-6 Astra with over 1M input tokens and five reasoning levels, where benchmark data on ARC AGI 3 shows higher per-token pricing can cut total task costs through efficiency gains. Anthropic updated Claude Fable 5.1 to match leading scores on the Artificial Analysis Intelligence Index v4.3 while easing cybersecurity limits and reducing verbosity. Google, Meta, and Microsoft pushed new speech-to-text models under 4% word error rates, and Johns Hopkins with Apple introduced SelfCompact for rubric-based LLM context pruning. These large language model updates accelerate AI industry impact and force AI engineering skills evolution toward fast feedback loops and high-agency ownership.