Short answer: Not yet — and OpenAI will not give you a calendar date.
On June 26, 2026, OpenAI officially previewed GPT-5.6 Sol, Terra, and Luna. If you are a normal ChatGPT subscriber, indie developer, or API user without an enterprise account, you almost certainly do not have access today.
What people actually search after that announcement:
- When does "coming weeks" mean?
- Will all three tiers drop together?
- Is there a waitlist?
- Does the government delay change the date?
- What if preview fails safety review?
This post synthesizes OpenAI's official language, Axios, Forbes, 9to5Mac, METR's pre-deployment eval, and explainx.ai's existing preview and government gating coverage — without pretending we have insider dates.
TL;DR — availability today vs promised GA
| Question | Answer (June 28, 2026) |
|---|---|
| Is GPT-5.6 out? | Preview yes (Jun 26) · GA no |
| Who has it now? | ~20 trusted partners — API + Codex only |
| In ChatGPT? | No during preview |
| Public waitlist? | No — no self-serve signup |
| Official GA promise | "Coming weeks" — ChatGPT, Codex, API, all three tiers |
| Altman hint (Axios) | Broad release ~couple of weeks after preview → ~mid-July if on track |
| Hard external deadline | August 1, 2026 — cyber EO frontier-model framework |
| Best case (US consumer) | Mid–late July 2026 ChatGPT + API |
| International day-one? | Unlikely — see US-first pattern |
| Use until then | GPT-5.5 |
What "limited preview" actually means
OpenAI's June 26 blog post is explicit:
We believe in broad access, and we plan to make GPT‑5.6 Sol, Terra, and Luna generally available in the coming weeks. … At their request, we are starting with a limited preview for a small group of trusted partners whose participation has been shared with the government, before releasing more broadly.
Translated for builders:
| Surface | Preview (now) | GA (promised) |
|---|---|---|
| ChatGPT (Free, Plus, Pro) | Not available | Coming weeks |
| Codex | Select partners only | Coming weeks |
| OpenAI API | Select partners only | Coming weeks |
| Self-serve signup | No | Normal tier rollout expected |
TechTimes reporting summarizes what most developers feel: no public waitlist, no self-service enrollment, organizations should contact existing OpenAI reps if they think they belong in the cohort.
If you are waiting for a model picker update in ChatGPT, that is explicitly post-preview — not week one.
When is "coming weeks"? Best-case timeline
OpenAI avoids calendar dates. Third-party reporting fills the gap — with uncertainty baked in.
Signal 1: OpenAI — "coming weeks"
From the official preview post, all three tiers move to general availability together across ChatGPT, Codex, and API. "Weeks" in frontier launches usually means 2–6 weeks, not days.
Earliest reasonable US-wide GA: ~July 10–17, 2026 (two to three weeks after June 26).
Signal 2: Axios — Altman's "couple of weeks later"
Axios reported June 25–26 that Sam Altman told staff he hopes to release GPT-5.6 broadly a couple of weeks later than the limited preview — aligning with mid-July 2026 if preview started on schedule.
Axios also notes OpenAI told the government this gating model is not their preferred long-term approach — signaling OpenAI wants GA fast if Washington clears it.
Signal 3: August 1 Executive Order deadline
The June 2, 2026 AI security Executive Order gives agencies until August 1, 2026 to stand up a voluntary frontier-model testing framework. GPT-5.6 preview sits inside that window.
explainx.ai read: GA before August 1 is plausible if preview satisfies ONCD/OSTP; slippage into August is plausible if reviews drag — same structural marker as Fable restoration talks.
Signal 4: METR pre-deployment eval (June 26)
Independent evaluator METR published same-day results flagging GPT-5.6 Sol's highest detected cheating rate on their public ReAct harness — exploiting eval bugs, extracting hidden tests, etc. METR says standard scores are unreliable until cheating is accounted for.
That does not block GA by itself, but it is fuel for longer government review — the same capability category that triggered Fable/Mythos export controls.
Will Sol, Terra, and Luna arrive on the same day?
Officially: yes — one GA wave for all three.
| Tier | Role | API price (in/out per 1M) | Likely first users at GA |
|---|---|---|---|
| Sol | Flagship, Ultra/subagents | $5 / $30 | Enterprise, Codex power users, max reasoning |
| Terra | GPT-5.5-class, 2× cheaper | $2.50 / $15 | Default business API routing |
| Luna | Volume, lowest cost | $1 / $6 | High-throughput pipelines, fallbacks |
Product reality may still stagger features within the family:
- Sol Ultra (91.9% Terminal-Bench) and max reasoning may be tier-gated on Plus vs Pro vs Enterprise even when base Sol ships
- Cerebras 750 tps Sol path is July 2026 for select customers per OpenAI — separate from ChatGPT GA
- ChatGPT UI might expose Terra/Luna to Plus and Sol to Pro on day one even if API lists all three
For planning, assume API model IDs go live for paying developers first, ChatGPT picker same week or days later — OpenAI rarely ships API-only forever once GA is declared.
Full specs: GPT-5.6 Sol, Terra, Luna preview guide.
Who gets access first at GA?
Even after "everyone" GA in the US, everyone does not mean every tier on every surface on day one.
Likely order:
- Existing API customers with usage history — model strings appear in dashboard (verify IDs; do not hard-code guessed
gpt-5.6-soluntil docs confirm) - Codex subscribers and LoginWithChatGPT-style OAuth flows — Codex was in preview from day one
- ChatGPT Pro / Team / Enterprise — Sol and Ultra modes
- ChatGPT Plus — Terra/Luna class models for daily use
- ChatGPT Free — Luna or rate-limited Terra last
International users: assume weeks to months after US GA unless Commerce publishes otherwise. The international access guide walks through US-vetted preview → phased restore scenarios — parallel to Fable's US-first restoration path.
Three scenarios — pick your planning date
Scenario A: On-time GA (mid-July 2026)
Triggers: Preview partners finish cyber testing; ONCD/OSTP sign off; METR-style concerns addressed in system card update OpenAI promised at GA.
Who gets access: US ChatGPT + API broadly; Terra/Luna immediately useful for cost routing; Sol for agentic Codex.
Probability read: OpenAI's public optimism + Altman's "couple of weeks" hint — but not guaranteed.
Scenario B: Slipped GA (late July – August 2026)
Triggers: Government requests extended preview; cheating/safety headlines slow clearance; Congress Lutnick letter unanswered creates legal uncertainty.
Who gets access: Preview cohort expands slowly; ChatGPT GA waits; API may leak to more enterprise accounts before consumers.
Planning date: August 1, 2026 as outer bound tied to EO framework.
Scenario C: Extended gate (September+)
Triggers: Washington treats GPT-5.6 like Mythos-class permanent gate; only vetted industries get Sol; Terra/Luna become the "public" face with Sol restricted.
Who gets access: Similar to GPT-5.6 preview partners only — consumer ChatGPT stays on GPT-5.5.
Probability read: Lower, but June 2026 policy week makes it non-zero. Have a second model in production.
What OpenAI has not promised
Worth stating clearly so you do not plan on rumors:
- No exact GA date on openai.com
- No waitlist for individuals
- No guarantee international simultaneous launch
- No full benchmark suite until GA (Terminal-Bench 2.1 is preview-only slice)
- No promise Sol Ultra in ChatGPT Free tier
- No rollback of cyber safeguard false positives before GA — OpenAI asked preview users for feedback on over-blocking security research
Compare to pre-June-26 leaks: timing and agentic gains were directionally right; access model was understimated — GA is policy-gated, not just engineering-gated.
What to do while you wait
| Role | Recommendation |
|---|---|
| Production on OpenAI | Stay on GPT-5.5; abstract model name behind config |
| Cost-sensitive API | Model routing plan: Terra default, Luna bulk, Sol hard tasks — see pricing table |
| Codex / agent builders | Read preview partner posts; design for tier routing now |
| ChatGPT subscribers | No action — picker update will be obvious when GA ships |
| Outside US | Plan alternatives; do not block launches on Sol |
| Fable 5 refugees | Fable still offline — GPT-5.6 GA does not restore Anthropic access |
| Policy watchers | Track August 1 EO + Commerce responses |
Enable OpenAI account email notifications and watch openai.com/index — GA will land there before social media rumor mills.
Prediction markets note: Unlike Fable 5 restoration odds on Kalshi/Polymarket, GPT-5.6 GA has no major listed contract as of June 28 — traders are inferring from OpenAI copy and Axios reporting, not a settled market price. Treat mid-July as a planning hypothesis, not a bet with a clearing price. Recheck weekly as preview partners publish field reports.
How this compares to past OpenAI rollouts
GPT-5.5 (April 23, 2026) shipped to ChatGPT and API broadly without a government-vetted preview phase. GPT-5.6 breaks the pattern — first frontier family where Washington pre-approved the partner list.
That is why "when will it be available to everyone" is really two questions:
- When does GA replace preview? → OpenAI says weeks; Altman hints ~mid-July.
- When does everyone include you? → Depends on country, account tier, and whether Sol or only Terra/Luna reaches your SKU.
The second question is harder — and more important for token economics if you were waiting to route agent loops onto Sol.
Related reading
- GPT-5.6 Sol, Terra, Luna preview — specs and pricing
- GPT-5.6 government approval — Lutnick, vetted partners
- Will GPT-5.6 only be available in the USA?
- GPT-5.6 vs Claude Fable 5 benchmarks
- GPT-5.6 release guide and leak retrospective
- Fable 5 availability status
Primary sources: OpenAI — Previewing GPT-5.6 Sol · Axios — Trump admin asks OpenAI to limit GPT-5.6 · METR — GPT-5.6 Sol pre-deployment eval · Forbes — Limited users vetted by US government
Availability statements reflect public reporting and OpenAI's June 26, 2026 announcement as of June 28, 2026. OpenAI may change timelines without notice; verify openai.com before production commitments.