a daily news desk
Deployments

OpenAI ships GPT-5.6 family after 12-day Commerce Department gate

Sol, Terra and Luna go public July 9 following a CAISI security review — and land as Microsoft 365 Copilot's 'preferred model' amid breakup chatter.

OpenAI made the GPT-5.6 family generally available Thursday, releasing Sol, Terra and Luna after a 12-day pre-release window in which roughly 20 vetted organizations, whose identities were shared directly with the U.S. government, ran the models behind closed doors. The Commerce Department’s Center for AI Standards and Innovation cleared the rollout after additional testing, per Axios reporting cited by Engadget, with OpenAI dispatching technical staff to Washington to field agency questions.

The gate is the first live stress test of the AI cybersecurity order President Trump signed in early June, which asks frontier labs to voluntarily submit models for review of up to 30 days. Twelve days is well under the ceiling. TechTimes reports that Commerce Secretary Howard Lutnick sought assurances from Sam Altman that every relevant agency had signed off before wider release, a choreography that resembles the pre-clearance dynamics around FDA advisory panels more than anything the software industry is used to.

The technical case is aggressive. Sol scores 80 on the Artificial Analysis Coding Agent Index, 2.8 points above Anthropic’s Fable 5, while using less than half the output tokens, less than half the time, and roughly a third less cost than its nearest competitor. On ExploitBench, Sol hits 73.5% against GPT-5.5’s 47.9% at the same token budget; on ExploitGym with a six-hour cap, it reaches 33.7%, more than double GPT-5.5’s 15.1% peak. A new “ultra” mode coordinates multiple agents across parallel workstreams. Altman told CNBC that Sol is 54% more token-efficient on coding than prior models. OpenAI says roughly 700,000 A100e GPU hours of automated red teaming produced safeguards that block about ten times more harmful activity than earlier Sol-class systems.

Pricing lands at $5/$30 per million input/output tokens for Sol, $2.50/$15 for Terra, and $1/$6 for Luna, with a Sol Fast tier on Cerebras hardware running up to 750 tokens/second at $12.5/$75, per Engadget.

The commercial subtext is louder than the technical one. OpenAI simultaneously announced GPT-5.6 as the “preferred model” inside Microsoft 365 Copilot, spanning Word, Excel, PowerPoint and Cowork, after Bloomberg reported Microsoft had been substituting in-house MAI models in some apps to cut costs. “Our partnership with Microsoft has always been about bringing the benefits of advanced AI to more individuals and organizations,” OpenAI said. Partnerships don’t usually need to be reaffirmed on launch day unless someone is counting the exits.

Sources