OpenAI launches GPT-5.6 after government review, with three tiers and ChatGPT Work

OpenAI has publicly released GPT-5.6, its latest AI model family, after weeks of government review over cybersecurity concerns. The rollout includes three variants—Sol, Terra, and Luna—and a new agent tool, ChatGPT Work, designed for enterprise productivity Source: The Verge.

Model Tiers and Pricing

GPT-5.6 is priced lower than its predecessor but remains more expensive than Chinese rivals. Sol costs $5 per million input tokens and $30 output, while Terra is $2.50/$15, and Luna is $1/$6 Source: SCMP. By contrast, Zhipu AI's GLM-5.2 charges about $1.40 input and $4.40 output, and DeepSeek V4 is even cheaper Source: SCMP.

• Performance claims: OpenAI says Sol scores 53.6 on Agents' Last Exam, beating Anthropic's Fable 5 by 13.1 points, and is 54% more token efficient on agentic coding Source: ZDNet, CNBC.
• ChatGPT Work: Combines ChatGPT and Codex for non-coding tasks, directly competing with Anthropic's Claude Cowork Source: The Verge.

Government Approval Controversy

The US government's role in approving the release is disputed. Reports from Axios and others said the Department of Commerce gave a "green light" after additional testing Source: Engadget. However, the White House denied providing any approval, stating no permission is required or granted Source: Gizmodo. The review involved the Center for AI Standards and Innovation, and OpenAI sent technical staff to Washington Source: TechCrunch. This process followed an executive order on voluntary pre-release checks, but critics note a lack of clear rules Source: The Guardian.

Competition and Market Impact

OpenAI directly targets Anthropic, which faced similar government restrictions on its Fable and Mythos models. Meanwhile, Chinese AI firms gain ground with cheaper, accessible models during US regulatory pauses Source: CNBC.