OpenAI’s new GPT‑5.6 “Sol” model is debuting as a breakthrough system for work and cybersecurity just weeks after government scrutiny delayed its release, sharpening a global race to automate knowledge jobs faster than regulators and workers can adapt.

A delayed debut after government review

OpenAI first unveiled the GPT‑5.6 family—flagship Sol, mid‑tier Terra, and budget Luna—in June, but full public access was postponed at the request of the Trump administration, which reviewed the system’s advanced cybersecurity capabilities.¹ The model supports defensive uses such as threat modeling, code review, and simulated attacks, and OpenAI bills it as its “strongest cybersecurity model yet, achieving frontier performance with significantly fewer tokens.”¹

On July 9, after receiving a government green light, OpenAI rolled out GPT‑5.6 publicly alongside ChatGPT Work, a new agentic assistant that integrates ChatGPT and Codex to “gather context from the apps, files, and workflows you choose and create finished materials such as documents, spreadsheets, presentations, and web apps.”²

From preview to office ‘super app’

The launch turned ChatGPT into what one report called “a super app for work,” merging the Codex coding tool, a new Work setting, and the GPT‑5.6 family into a single desktop hub for engineers and other professionals.³ Mac and Windows users, including many on free tiers, were promised rapid access as the rollout “starting globally” moved toward “full availability over the next 24 hours.”²

At the same time, Axios noted a new “paradox of choice”: non‑technical users must now pick between Sol, Terra, and Luna, plus varying “reasoning” levels, to find the “just right” balance of speed, cost, and accuracy for each task.⁴

Hype, benchmarks, and infrastructure strain

OpenAI executives and partners quickly touted frontier results. TechCrunch reported that CEO Sam Altman claims Sol is 54% more token‑efficient for coding and that GPT‑5.6 outperforms rival Anthropic on key coding benchmarks.¹ A system card from OpenAI’s safety team described Sol as the flagship, Terra as a lower‑cost option, and Luna as the fastest and most efficient model, emphasizing that the launch uses the company’s “most robust yet” safeguards for safe global deployment.⁵

External platforms raced to integrate the models. Amazon announced that Sol, Terra, and Luna are “now generally available on Amazon Bedrock” as three tiers “from flagship reasoning to fast inference,” drawing an enthusiastic “GPT‑5.6 is now generally available on Bedrock:” from OpenAI president Greg Brockman.Greg Brockman (gdb)⁶ Microsoft CEO Satya Nadella highlighted GPT‑5.6 with “Work IQ” arriving in Copilot Chat, Cowork, M365 apps, GitHub, and Foundry, promising “stronger reasoning and higher-quality outputs without sacrificing efficiency.”Satya Nadella (satyanadella)⁷

Benchmarks and design contests reinforced the hype. A tweet amplified by Brockman celebrated Sol ranking “1st overall on Design Arena with an Elo of 1353,” edging out Anthropic’s Claude Fable 5 on frontend design.Greg Brockman (gdb)⁸ Another viral evaluation from ARC Prize, shared by Sam Altman, claimed “GPT‑5.6 Sol sets a new SOTA on ARC‑AGI‑3… the best model at orienting in a situation it’s never encountered.”Sam Altman (sama)⁹

Users test the limits—and Altman warns of “hiccups”

Early adopters and competitors weighed in. An Axios columnist framed the new Work agent as a powerful but complex tool that “can use connected apps and files to research, analyze information and create documents, spreadsheets, presentations, reports and websites,” while warning that workers must learn to tune models and reasoning levels to avoid cost overruns or poor results.⁴ Another piece described OpenAI’s strategy as “train uber-powerful AI models, deliver them to ChatGPT’s massive user base, and become an indispensable part of life for millions of workers.”³

On X, one user quoted by Sam Altman admitted many posts about missing rival model Fable were “actually about missing GPT 5.6 Sol,” concluding, “Fable is really good and maybe better in some ways, but I vastly prefer GPT 5.6 Sol.”Sam Altman (sama)¹⁰ Another reviewer, amplified by Brockman, called GPT‑5.6 Sol “the most impressive model ever” with “insanely good” visual math performance and “huge improvements” over other systems.Greg Brockman (gdb)¹¹

The enthusiasm has created its own bottlenecks. On July 14, Altman cautioned that “5.6 sol growth is insane” and, despite “heroic work” by the inference team, “it is possible there are some hiccups soon.”⁶ In another post, he leaned into the hype, saying there are “a lot of benchmarks that suggest 5.6 sol is the best model in the world right now,” joking that “the most reliable way to tell is that elon is obsessed with me again.”⁶

Safety systems race to keep pace

Behind the scenes, OpenAI’s AI‑authored GPT‑5.6 System Card presents the company’s own perspective on risk and governance. It describes a three‑model family—Sol, Terra, Luna—and stresses that the “safeguards we have built for this launch—our most robust yet—are built to deliver these models safely and at scale, around the world.”⁵

Yet human commentators note that these advances are arriving faster than institutions can absorb them. One analysis argued that companies like OpenAI and Anthropic are locked in a race to build AI agents “actually useful for the average person,” even as the “theoretical right-hand AI agent for the everyday consumer remains out of reach.”² Business coverage framed GPT‑5.6 and ChatGPT Work as OpenAI’s “biggest play for the office,” part of a broader effort to dominate enterprise workflows just as the company eyes a potential public offering.³

As GPT‑5.6 Sol spreads across cloud platforms and productivity suites, the central tension is no longer whether such systems will reshape work, but whether governments, infrastructure, and workers can keep up with a model its creator calls both the “best” yet—and one that may already be straining the world built to contain it.