OpenAI Unveils GPT-5.6 at Half Anthropic's Price
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- OpenAI unveiled GPT-5.6 — a suite of three models: Sol (flagship), Terra (medium-tier for "high-volume work"), and Luna ("fast and affordable") — less than 24 hours after the Trump administration requested a staggered release.
- Sol is priced at $5 input/$30 output per million tokens, roughly half the cost of Anthropic's Claude Fable 5 ($10/$50), with Terra and Luna each priced at less than half the tier above them.
- OpenAI debuted two additional Sol modes — "max" for deeper reasoning and "ultra" for sub-agents — the latter evoking OpenClaw and potentially reflecting creator Peter Steinberger's work at the company.
- OpenAI dedicated approximately 700,000 A100e GPU hours to automated red-teaming; third-party testers will continue testing for two weeks, and Sol does not cross the cyber-critical threshold under the company's preparedness framework (revised in April with some areas of previous study removed).
- The announcement appeared to reference Anthropic's recent jailbreaking incidents, with OpenAI stating GPT-5.6 is "trained to refuse prohibited cyber assistance, including when users attempt to disguise their intent or jailbreak the model."
- The Trump administration will approve OpenAI customers case-by-case during the preview period; OpenAI said it cooperated "hoping it wouldn't be the norm" and warned that this kind of government access process should not become "the long-term default."
Why it matters: The Trump administration now has case-by-case approval power over who can access OpenAI's frontier model during the preview — an unusual government gatekeeping role for a commercial AI release — while Sol undercuts Anthropic's Claude Fable 5 by ~50% on input/output pricing, intensifying the commercial AI race even as regulatory friction grows.
