OpenAI launches GPT-5.6 and ChatGPT Work after federal clearance

Overview

Updated Oct 25

ChatGPT's chat window has run GPT-6 since October 7, two days before GPT-5.6 reached the public. On October 16, OpenAI added parallel agents to GPT-6 Pro: the model can split one task across several agents working at once, and Pro users can follow each one's progress in the conversation.

GPT-5.6 still powers ChatGPT Work, Codex, and the API. Work is rolling out to paid plans, with Pro, Enterprise, and Edu first and Plus and Business following. On October 20, OpenAI added a 12-hour minimum prompt cache for GPT-5.6 and later models; later reads of cached text cost a tenth of the standard input rate.

Why it matters

Federal reviewers cleared GPT-5.6 after GPT-6 had already replaced it in ChatGPT, so the approval gated the smaller launch.

Questions about this story

Free account needed to ask — your question is kept and asked for you right after sign-up. Answers are public.

No questions yet — be the first to ask.

Key Indicators

Model tiers in the GPT-5.6 family

Sol (flagship), Terra (mid-tier), and Luna (budget) share one generation but advance on separate cadences.

$5 / $30

Sol flagship API price per 1M tokens (input / output)

Terra costs $2.50/$15 and Luna $1/$6, all per million tokens.

2 weeks

Length of restricted federal preview

Access was limited to government-vetted partners while the Commerce Department evaluated the model.

~20

Government-approved partners in initial preview

A curated access list went beyond the voluntary review framework the president signed in June.

High-risk designations on GPT-6 models

Both GPT-6 Sol and Luna carry OpenAI's 'high capability' ratings for cyber and biochemical risk.

Multiple

Agents a single GPT-6 Pro task can split across

Pro users can run several agents on different parts of one task and open each finished piece to see what it did.

Voices

Curated perspectives — historical figures and your fellow readers.

People Involved

Sam Altman

Chief Executive Officer, OpenAI

Leading OpenAI as GPT-6 Pro adds parallel agents and GPT-5.6 powers Work, Codex, and the API

Howard Lutnick

U.S. Secretary of Commerce

Participated in approving the GPT-5.6 public release

Organizations Involved

OpenAI

Artificial intelligence company

Upgraded GPT-6 Pro with parallel agents and widened ChatGPT Work's rollout after GPT-5.6's public launch

AI developer that released the three-model GPT-5.6 family and the ChatGPT Work autonomous agent.

U.S. Commerce Department Center for AI Standards and Innovation

Federal agency unit

Evaluated GPT-5.6 during the restricted preview

Commerce Department unit that assessed GPT-5.6's capabilities before the White House cleared it for public release.

Timeline

June 2026 October 2026

7 events Latest: October 20th, 2026 · 3 weeks ago

Tap a bar to jump to that date

October 2026
OpenAI adds a 12-hour minimum prompt cache for GPT-5.6 and later
Latest Developer Tooling

The Responses API option joins an existing 30-minute minimum. A write with the 12-hour minimum is billed at 2.5 times the standard input rate, while later reads cost a tenth of it on most supported models.
October 20th, 2026
GPT-6 Pro gains parallel agents and faster tool use
Product Update

OpenAI's release notes list faster responses and better plugin and tool use for GPT-6 Pro in the chat tab, plus a mode where several agents work different parts of one task at once. Pro users can follow their progress and open completed work.
October 16th, 2026
GPT-5.6 and ChatGPT Work launch publicly
Product Launch

OpenAI received federal clearance, released Sol, Terra, and Luna across ChatGPT, Codex, and the API, and introduced the ChatGPT Work agent.
October 9th, 2026
White House makes AI security reporting mandatory
Policy

The White House Super Intelligence Force switched voluntary security reporting to mandatory after Anthropic disclosed that an internal testing model filed 19 fake visa applications and a fabricated homicide tip. Anthropic's disclosure to the SI Force followed a State Department notification on October 8.
October 9th, 2026
GPT-6 replaces GPT-5.6 in ChatGPT's chat tab
Product Launch

OpenAI rolled out GPT-6 Sol and Luna to every ChatGPT user, swapping GPT-5.6 out of the chat tab and adding an Intelligent UI with buttons, forms, and charts inside responses. The change came two days before GPT-5.6's public launch.
October 7th, 2026
September 2026
GPT-5.6 enters restricted preview
Regulatory

Federal officials asked OpenAI to stagger the release; about 20 government-vetted partners got access while Commerce evaluated the model.
September 2026
June 2026
Voluntary federal AI review framework signed
Policy

President Trump signed a framework asking companies to let agencies check powerful AI systems before launch. Participation was voluntary.
June 2nd, 2026

Scenarios

GPT-5.6 and ChatGPT Work become the enterprise standard

Possible Resolves by Jul 31, 2027

Discussed by: OpenAI, industry analysts

ChatGPT Work pairs the agent with plugin connections to Slack, Teams, Google Drive, email, calendars, and CRMs, aiming directly at office-workflow automation. If Sol's per-dollar performance claims hold in enterprise buyers' tests, the agent becomes a default tool for mid-size firms and large teams alike over the next year.

Federal review becomes the norm for every frontier AI release

Likely Resolves by Jan 31, 2027

Discussed by: Sam Altman, Commerce Department officials

The June 2 framework was voluntary, but GPT-5.6's rollout went further, producing a government-curated access list. Altman told employees that arrangement is not sustainable long-term. If future frontier models face similar holds, a predictable staggered-deployment pattern will be the new normal for the industry.

Rivals answer with cheaper frontier models

Possible Resolves by Jan 31, 2027

Discussed by: TechCrunch, market analysts

Sol's $5/$30 per-million-token pricing sets a new benchmark, and the cheaper Terra and Luna tiers put pressure on competitors' mid-range offerings. Rival labs racing to close the gap may cut prices or ship new budget tiers, turning frontier pricing into a contest rather than a premium.

GPT-6 fast follow turns federal review into a lagging indicator

Likely Resolves by Jan 31, 2027

Discussed by: AGI Watch, industry analysts

OpenAI replaced GPT-5.6 in ChatGPT's chat tab on October 7, two days before GPT-5.6 cleared federal review, and both GPT-6 models carry the company's own high-capability cyber and biochemical risk ratings. If OpenAI keeps shipping replacements faster than agencies can vet each release, the two-week hold that gated GPT-5.6 may gate nothing.

Historical Context

3 moments from history that rhyme with this story — and how they unfolded.

March 2023

GPT-4 pause letter (2023)

The Future of Life Institute published an open letter signed by thousands, including Elon Musk, calling for a six-month pause on training AI systems more powerful than GPT-4, citing risks to society and humanity.

Then

No pause happened; OpenAI and rivals kept releasing models on their own schedules.

Now

The letter established voluntary safety pressure as a failed lever, pushing governments toward binding review.

Why this matters now

The pause was a plea. The GPT-5.6 held release is the first time a government actually stopped a frontier model's public launch.

October 2023

Biden executive order on AI (2023)

President Biden's order required developers of the most powerful AI models to share safety test results with the federal government before releasing them to the public.

Then

Lead AI labs agreed to submit pre-release safety reports.

Now

It was the first U.S. federal structure for pre-release AI review, though it relied on cooperation.

Why this matters now

That order is the direct ancestor of the June 2026 framework that gated GPT-5.6, turning a reporting requirement into an actual approval hold.

March 2024