GPT-6 Astra and the three models that make up GPT-6
The model id is gpt-6-astra[1]. OpenAI describes it as its most capable model, built for the hardest end-to-end work[1,2].
The GPT-6 generation became three models on 22 September 2026, when GPT-6 Sol and GPT-6 Luna were added[3,4]. Per million tokens, Astra is $10 input and $50 output, exactly five times Sol and a hundred times Luna[4].
The context window is 1,050,000 tokens and the maximum output in one response is 128,000 tokens[1]. Those are the same figures the GPT-5.6 models carry, so the generation did not lengthen either one[2].
The reliable knowledge cutoff is 30 April 2026[1]. The model produces reasoning tokens[1].
The GPT-5.6 line is still offered[4]. It is gpt-5.6-sol, which also answers to gpt-5.6, then gpt-5.6-terra, gpt-5.6-luna, and gpt-5.6-cyber, which is for security work, and all four hold the same 1,050,000 tokens as Astra[2].