GPT-5.6 availability by plan: Sol, Terra and Luna

Where GPT-5.6 is available today across ChatGPT, Work, Codex and API

OpenAI switched on all three GPT-5.6 models for general use on July 9, 2026, in ChatGPT, Codex and the OpenAI API on the same day. The rollout started worldwide at once, and OpenAI expected it to reach full availability within 24 hours.

The family has three tiers. Sol is the flagship, Terra costs less and is described by OpenAI as competitive with GPT-5.5 on performance, and Luna is the fastest and cheapest. The 5.6 names the generation, and OpenAI says each tier name can move to a newer model on its own schedule.

What you can reach depends on where you use it. The table maps each product to the three tiers, based on the OpenAI Help Center page as saved on September 1, 2026.

Which GPT-5.6 tiers each product offers (OpenAI Help Center, September 1, 2026)

Product	Sol	Terra	Luna
ChatGPT chat	Plus, Pro, Business, Enterprise	Not selectable	Default model on Free and Go, not selectable
Work in ChatGPT	Plus, Pro, Business, Enterprise	Plus, Pro, Business, Enterprise	Plus, Pro, Business, Enterprise
Codex	Plus, Pro, Business, Enterprise	Free, Go, Plus, Pro, Business, Enterprise	Plus, Pro, Business, Enterprise
OpenAI API	Available	Available	Available

Free and Go users therefore meet a different GPT-5.6 model in each product: Luna in chat and Terra in Codex. Sol requires Plus or a higher plan everywhere except the API, where it is billed per token.

Which ChatGPT plan gets which reasoning level

Standard ChatGPT chat is where most confusion starts, because you choose a reasoning level there rather than picking Sol, Terra or Luna by name. Your plan decides which levels appear.

Standard chat: Sol only, chosen by reasoning effort

Reasoning effort is how much work the model does on a request before it replies, and ChatGPT offers five levels of it. Instant gives quick everyday answers, while Medium and High add standard and extended reasoning. Extra High is the highest effort ChatGPT offers with Sol, and Pro is for difficult, longer-running tasks.

On eligible paid plans, Instant through Extra High all run on GPT-5.6 Sol, and Pro runs on GPT-5.6 Sol Pro. Plus adds Medium and High to Instant, while Pro, Business and Enterprise get every level.

GPT-5.6 Sol reasoning levels by ChatGPT plan (OpenAI Help Center, September 1, 2026)

Plan	Medium and High	Extra High	Pro (Sol Pro)
Plus	Included	Not included	Not included
Pro	Included	Included	Included
Business	Included	Included	Included
Enterprise	Included	Included	Included
Free and Go	Not included	Not included	Not included

Free and Go have no Sol at any level. Their default model is GPT-5.6 Luna, and the Think button they get for harder questions also runs on Luna. Logged-out visitors cannot use Sol at all.

On eligible paid plans, ChatGPT can raise the reasoning level on its own for a complex request while Instant stays selected. You turn that off or on under Settings > General > Higher intelligence. In Business and Enterprise workspaces, admins can limit which models members can use, and Enterprise workspaces may keep the old model picker until the new controls reach them.

At launch, GPT-5.5 Instant was still the default in chat. That changed on August 6, 2026, when OpenAI moved Plus and Pro to a chat-tuned Sol for both Instant and deeper reasoning, and said Luna would become the Free and Go default that week. The Sol in Work and Codex was left unchanged, so the same name now covers two versions.

Two OpenAI pages also disagree on Sol Pro. The July 9 announcement lists it for Pro and Enterprise only, while the Help Center table saved on September 1 includes Business as well, and this article follows the Help Center. If a level you expect is missing, the rollout may not have reached your account yet.

The Help Center page has changed since launch, and the change was easy to miss. Keeping a Markdown copy of it lets you compare the next version line by line.

Terra and Luna live in Work, Codex and the API

Terra and Luna cannot be picked in a standard ChatGPT conversation. You reach them in three other places: Work in ChatGPT, Codex and the OpenAI API.

Work in ChatGPT offers all three tiers to Plus, Pro, Business and Enterprise. Codex gives Free and Go users Terra and gives paid plans all three, but it needs version 26.707.30751 or later of the ChatGPT desktop app (Codex mode), or version 0.144.0 or later of the Codex CLI. In both products you set an effort level per model, and the max effort level is open to anyone with GPT-5.6 access once it is switched on in settings.

Ultra is a separate mode that splits a large job across subagents. OpenAI limits it to Pro and Enterprise in Work, and to Plus and higher plans in Codex.

In the API each tier has its own model ID, for example gpt-5.6-terra, and reasoning effort can be set anywhere from none to max. The shorter gpt-5.6 alias points to Sol, so a request that names only the generation is billed at Sol's rate.

GPT-5.6 has no country list of its own. ChatGPT access follows ChatGPT's supported countries, and API access follows the API's. The Help Center names the EEA, Switzerland, the United Kingdom and the UAE as covered, but workloads set to UAE inference residency cannot currently use GPT-5.6.

What GPT-5.6 costs, tier by tier

In ChatGPT, GPT-5.6 comes with your plan. In the API you pay per million tokens, the units the API uses to count text, with separate rates for the input you send and the output you get back.

All three tiers are cheaper than at launch. On July 30, 2026, OpenAI cut Luna by 80% and Terra by 20%. On August 21, it cut Sol's API and credit pricing by over 20% for three months, and the API docs say that promotional rate lasts at least through November 21, 2026.

GPT-5.6 API price per 1M tokens, input / output: at the July 9 launch and now

Tier	July 9, 2026 launch	Now (pricing page, September 25, 2026)
Sol	$5 / $30	$4 / $20 (promotional)
Terra	$2.50 / $15	$2 / $12
Luna	$1 / $6	$0.20 / $1.20

For scale, GPT-5.5 costs $5 input and $30 output, and GPT-5.4 costs $2.50 and $15. Terra now charges less than GPT-5.4 on both, even though OpenAI compares it with GPT-5.5, the pricier model. For choosing a tier by workload, see GPT-5.6 Sol, Terra and Luna: Specs and Pricing.

Standard and batch pricing across Sol, Terra and Luna

The pricing page lists two rates that matter for most API users. Standard is the regular rate, and Batch is the rate for requests sent through the Batch API endpoint. Cached input is the cheaper rate for the part of a prompt the API has already stored from an earlier request.

GPT-5.6 API pricing per 1M tokens, under 272K input tokens (OpenAI pricing page, September 25, 2026)

Tier	Standard: input / cached / output	Batch: input / cached / output
Sol (promotional)	$4 / $0.40 / $20	$2 / $0.20 / $10
Terra	$2 / $0.20 / $12	$1 / $0.10 / $6
Luna	$0.20 / $0.02 / $1.20	$0.10 / $0.01 / $0.60

Every Batch figure is half the Standard figure beside it, a ratio we worked out from the table rather than one OpenAI states. Writing a prompt into the cache costs 1.25x the uncached input rate, and reading it back keeps the 90% cached-input discount.

The 272K-token pricing cliff

All three tiers accept up to 922,000 input tokens inside a 1,050,000-token context window. Past 272K input tokens in a single request, though, the price changes for the whole request, not only for the tokens over the threshold.

That is why it works like a cliff. Cross 272K by a single token and every input token in the request costs 2x, while every output token costs 1.5x. Trimming a prompt back under the threshold, or splitting it into smaller requests, keeps it on the standard rate.

GPT-5.6 API pricing per 1M tokens above 272K input tokens (OpenAI pricing page, September 25, 2026)

Tier	Input	Cached input	Output
Sol (promotional)	$8	$0.80	$30
Terra	$4	$0.40	$18
Luna	$0.40	$0.04	$1.80

Why it wasn't available to everyone at launch

Everything above describes GPT-5.6 since July 9, 2026. From the June 26 announcement until general availability, only a small group of partners could use it.

A government-requested limited preview

On June 26, 2026, OpenAI opened GPT-5.6 only to a select group of trusted partners and organizations, through the API and Codex. ChatGPT users had no way in.

OpenAI said the US government asked for this. It had shown the government its plans and the models' capabilities before launch, and the government requested a preview limited to partners whose participation had been shared with it. OpenAI also said it did not want this kind of government access process to become the long-term default, and it moved to general availability on July 9.

Executive Order 14409, in plain terms

OpenAI's preview post names no order number. It says only that OpenAI was helping the Administration build what it calls "the cyber Executive Order framework", along with a process it can reuse for later launches.

At the time, the order in force that described this kind of arrangement was Executive Order 14409, signed on June 2, 2026. Section 3 tells agencies to design a voluntary framework in which AI developers could give the federal government access to a covered frontier model for up to 30 days before releasing it to other trusted partners. The same section says nothing in it authorizes a mandatory licensing, preclearance or permitting requirement for developing or releasing AI models.

OpenAI does not say whether its preview ran under this framework. Either way, the preview ended on July 9, 2026, so if GPT-5.6 is still missing for you today, look at your plan, your account or workspace settings, and the product you are using. For how ChatGPT's plans and model picker work in general, see GPT-5 in ChatGPT: Free Tier, Pricing, Model Picks.