OpenAI released GPT-5.6 in three sizes — Sol, Terra and Luna — priced at $5/$30, $2.50/$15 and $1/$6 per million input/output tokens, against Anthropic's Opus at $5/$25 and Fable at $10/$50. The launch post leads with Agents' Last Exam, the AA Agent Coding Index v1.1 and BrowseComp, and states that internal researchers' average daily output tokens per active researcher more than doubled versus GPT-5.5, with research compute for internal coding inference up 100-fold over six months. OpenAI also said Sol Ultra produced a Lean-formalized proof of the 50-year-old Cycle Double Cover Conjecture using 64 subagents in under an hour, open-sourced at openai/cdc-lean. In aggregator commentary, Sol scores 58.9 on the Artificial Analysis Intelligence Index, a notch behind Fable, at $1.04 versus $2.75 cost per task, and tops WeirdML at 88.8%.