OpenAI releases GPT-5.6 models Sol, Terra and Luna for general use
For the cto of a regional bank managing ai spend
OpenAI said on July 9 that it is launching the GPT-5.6 family for general availability after a limited preview, with Sol as the flagship, Terra for everyday work and Luna as the most cost-efficient model; the post names no location.1
OpenAI said in its launch post that it trained GPT-5.6 to get more useful work from every token.1 The company said Sol outperforms previous and competing frontier models with fewer tokens and at lower estimated cost, which it described as more successful work for the same spend or comparable results at lower total cost.1
OpenAI said its new `ultra` setting coordinates four agents in parallel by default, and that it trades higher token use for stronger results and faster time-to-result on demanding tasks.1
OpenAI added updates to the post after launch: on July 30 it reduced the price of Luna by 80% and Terra by 20%, and on Aug. 21 it dropped the API and credit pricing of Sol by over 20% for the next 3 months.1
On Agents' Last Exam, OpenAI reported that Sol scored 53.6, which it said is 13.1 points above Claude Fable 5 (adaptive reasoning); these are OpenAI's own reported results.1
Amazon Web Services said in a blog post that GPT-5.6 Sol, Terra and Luna are available through Amazon Bedrock cross-Region inference from the Canada (Central) and Canada West (Calgary) Regions. According to the post, the US geographic profile routes requests to the Canadian source Region and US Regions, and AWS says to choose it when inference must stay within North America. The global profile routes to supported commercial AWS Regions and, for GPT-5.6, is available only from ca-central-1.2
The per-task savings OpenAI describes apply to its efficient default, while OpenAI says ultra uses more tokens, so banks should test cost per task in each mode before budgeting. Ultra also runs several agents at once, which gives model-risk and validation teams more than one line of reasoning to review per answer.
OpenAI said on Aug. 21 that the Sol price cut lasts for the next 3 months and did not give an end date in the post.1
Our take. OpenAI says Sol does the work with fewer tokens, then offers an 'ultra' mode that runs several agents at once. The bill may shrink per task and grow per click, and model risk gets several reasoning trails instead of one.
Action. Model risk: Require logs from every parallel agent before approving GPT-5.6 'ultra' mode for any production or customer-facing use.
What our records say
ChatGPT Enterprise's posture, per its profile, is that it sells banks ChatGPT seats for employees and API access to its models, and in 2026 added on-site engineering deals, a financial-services edition built with investment-bank design partners, and a separately funded deployment company.1 Strategy and products: The product "ChatGPT Enterprise" is recorded with status Generally available.2
From our records
- ChatGPT Enterprise (OpenAI)vendor · banks using it (14) Morgan Stanley, BNY, Customers Bank · alternatives Microsoft Copilot, Claude, Gemini Enterprise
- Regions Financialbank · latest Core deposits planned for 2027 (Oct 2026) · vendors GitHub Copilot · enforcement 2 actions since 2016; latest Federal Reserve civil money penalty, $2.9 million (Aug 2023)
- Claude (Anthropic)vendor · banks using it (21) Goldman, JPMorgan, BankUnited · alternatives Microsoft Copilot, ChatGPT Enterprise, Gemini Enterprise
- Amazon Bedrock (AWS)vendor · banks using it (9) BankUnited, FNBO, First Interstate · alternatives Microsoft Copilot, ChatGPT Enterprise, Claude