OpenAI Launches GPT-5.6 Sol, Gates Access via US Government
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- OpenAI began a limited preview of GPT-5.6 Sol, its flagship model that matches Mythos Preview on the ExploitBench benchmark, alongside the balanced Terra model for everyday work and the fast, affordable Luna model.
- GPT-5.6 Sol introduces an Ultra mode with subagents for complex workflows and a max reasoning setting designed for deep problem-solving.
- OpenAI is restricting GPT-5.6 access at the US government's request, limiting the rollout to vetted preview partners rather than making it publicly available.
- OpenAI stated the restrictions "shouldn't be the norm," framing the limited rollout as an exception rather than standard policy going forward.
Why it matters: OpenAI's strongest model yet is being placed behind a US government approval gate — and the company's own framing that this "shouldn't be the norm" signals an unresolved tension between Silicon Valley's release cadence and Washington's preference for controlled frontier AI deployment. For the first time, the most powerful commercial AI release is gated by federal vetting rather than market access alone.
