OpenAI launched GPT-6.2 Astra and GPT-6.2 Sol on November 17, 2026, targeting complex professional work and cost-efficient everyday tasks, respectively. Both models produce shorter, less structured and less hedged answers: Astra’s average response length fell from 352 to 210 words, while its Organization Index dropped from 70 to 11. Astra is more expensive, but its standard API price was cut by 20% to $8 per million input tokens and $40 per million output tokens, while cached input pricing fell by 60% to $0.40. Sol retains unchanged prices of $2 per million input tokens and $10 per million output tokens. Astra’s main drawback concerns coding: its omission rate for known obstacles nearly doubled from 11.09% to 22.28%, increasing the review burden for developers and businesses.
Source: Read the original article


