The wait is finally over for AI enthusiasts and developers. After spending some time locked behind a restricted preview mandated by the Trump administration, OpenAI has officially launched its highly anticipated GPT-5.6 AI model family for general availability.
Speaking with CNBC, OpenAI CEO Sam Altman characterized the approval process as a collaborative “back and forth” with top government officials. The list includes Commerce Secretary Howard Lutnick and National Cyber Director Sean Cairncross. Federal reviewers closely scrutinized the model’s advanced capabilities in coding, biology, and cybersecurity before granting permission for a public release.
Three flavors of intelligence
Rather than dropping a single monolithic system, OpenAI is launching three distinct versions of GPT-5.6 simultaneously. Each one is tailored to different budgets and performance needs.
First is Sol, the premium flagship model for heavy reasoning and complex workflows. Next is Terra, a balanced AI for everyday office routine. Then there is Luna, the leanest of them all, designed for the fastest processing speed and high-volume operations.
The pricing strategy targets big companies. API access for the flagship Sol model sits at $5 per million input tokens and $30 per million output tokens. Meanwhile, the balanced Terra model costs $2.50/$15, and the budget-friendly Luna runs at just $1/$6. According to Altman, Sol achieves an impressive 54% greater token efficiency on programming tasks, directly addressing corporate concerns regarding spiraling AI operational costs.
Crushing benchmarks and delegating tasks
In terms of pure performance, OpenAI’s internal metrics show that GPT-5.6 Sol is taking direct aim at Anthropic’s current heavyweight, Fable 5. On “Agents’ Last Exam”—a rigorous test evaluating professional workflows across 55 fields—Sol outpaced Fable 5 by over 13 points while running at roughly a quarter of the estimated cost.
To achieve these numbers, OpenAI is introducing an “Ultra” setting for premium subscribers. When toggled, Ultra mode automatically coordinates four separate sub-agents to work on a demanding task in parallel. It trades higher token consumption for incredibly fast completion times and unparalleled accuracy.
Meet your new AI coworker
Alongside the core architecture update, the company also launched ChatGPT Work. This is a supercharged digital assistant designed to compete directly with Anthropic’s Claude Cowork. Powered by the new 5.6 infrastructure, the tool automatically gathers context across your connected files and third-party applications. It can build spreadsheets, format presentations, and draft documents across web, mobile, and desktop systems, rolling out first to Windows and Mac applications.
The Android Headlines Take
Strategically, splitting the GPT-5.6 update into Sol, Terra, and Luna is a smart move by OpenAI. The Google-like tiered approach allows the public accustomed to Gemini AI models to have a clearer understanding of what each model does, potentially attracting a segment of Google’s market. Furthermore, offering comparable flagship intelligence to Anthropic’s best silicon at a massive financial discount, while providing a hyper-efficient parallel agent system, is a strong bet for professionals and businesses looking to reduce their AI costs.
The post OpenAI Unleashes GPT-5.6 to the Public Following a Strict Government Review appeared first on Android Headlines.
