OpenAI just released GPT-5.6 after an unusual government-reviewed rollout — a three-tier family built for coding, cybersecurity, and everyday work. Here's what actually changed.
GPT-5.6 Just Launched: Sol, Terra, and Luna Explained
GPT-5.6 went live for everyone on July 9, 2026, but the way it got there was more unusual than the model itself. OpenAI did not just ship it: the release was delayed by a first-of-its-kind U.S. government safety review before the public ever got access. This guide explains why it was delayed, what each of the three models is for, how pricing and ChatGPT access have changed since launch, and how to decide which one to use.
Why the Delay Happened
OpenAI first previewed GPT-5.6 on June 26, but instead of a normal public rollout, access was limited to around 20 government-approved partner organizations. The reason was cybersecurity: GPT-5.6's strongest model showed frontier-level capability in vulnerability research, which was serious enough that the U.S. government asked OpenAI to hold broad release until a safety review cleared it. That review moved faster than expected, and general availability opened up on July 9.
Meet the Three Models
GPT-5.6 is not a single model. It is a family of three, each aimed at a different budget and use case.
Sol: the flagship
Sol is OpenAI's most capable model to date, tuned for coding, scientific reasoning, and cybersecurity work. It introduces a new "ultra" mode that can coordinate multiple sub-agents working in parallel to finish complex, multi-step tasks faster than a single model working alone.
Terra: the balanced option
Terra sits in the middle, aiming for everyday production work at roughly half the cost of the previous generation's comparable model, without a major drop in quality.
Luna: the budget workhorse
Luna is built for high-volume, simple tasks where cost per request matters more than squeezing out maximum intelligence.
Which Tier Should You Use?
| Task | Start with | Why |
|---|---|---|
| Classifying, tagging, or extracting data at scale | Luna | Lowest cost per request |
| Customer replies, summaries, drafting | Terra | Good quality at a mid-range price |
| Hard coding bugs, multi-file changes, security research | Sol | Strongest reasoning in the family |
| Long autonomous agent runs | Sol | Its "ultra" mode coordinates parallel sub-agents |
How It Performs
On coding-specific benchmarks, Sol set a new state-of-the-art score while using under half the output tokens and roughly a third of the cost compared with Anthropic's top coding model at the time. Independent early testers were split on which model "feels" smarter day to day, with some preferring Anthropic's model for raw creative reasoning and others favoring GPT-5.6 for consistency on everyday, practical tasks. If coding is your main use, compare both in our Claude vs ChatGPT comparison.
Pricing Breakdown
| Model | Input (per 1M tokens) | Output (per 1M tokens) | Best for |
|---|---|---|---|
| Sol | $5 | $30 | Hardest, most complex tasks |
| Terra | $2.50 | $15 | Everyday production work |
| Luna | $1 | $6 | High-volume, simple tasks |
Note: these are launch-day prices. Terra and Luna were cut on July 30, and Sol was cut on August 21 -- see the full updated pricing breakdown for current rates.
What Has Happened Since Launch
| Date (2026) | Event |
|---|---|
| June 26 | Limited preview to about 20 approved partner organizations |
| July 9 | General availability of Sol, Terra, and Luna |
| July 30 | Luna cut 80% and Terra cut 20% on the API |
| August 6-10 | Luna becomes the default for ChatGPT Free and Go, with unlimited text chats |
| August 21 | Sol API price cut by more than 20% |
The free-plan change is covered in detail in our guide to ChatGPT's unlimited free text chats.
New Alongside the Launch: ChatGPT Work
OpenAI also introduced ChatGPT Work, an agent-style tool built for longer, multi-step workplace tasks: pulling context from your existing tools to turn scattered notes and drafts into finished documents, spreadsheets, or presentations. It rolled out first on desktop, with mobile and web access following shortly after.
Where You Can Access GPT-5.6 Today
In ChatGPT, Luna is now the default model for Free and Go accounts, and free users get a "Think" button for harder questions. Plus and Pro subscribers can use an updated Sol that was retuned for everyday chat, with a slider that controls how much thinking goes into each answer. Terra and Luna are also accessible through ChatGPT Work, Codex, and the OpenAI API, depending on your plan. GPT-5.6 is also becoming the preferred model inside Microsoft 365 Copilot across Word, Excel, and PowerPoint.
Should You Switch?
If you are currently on GPT-5.5 or a comparable model, there is no urgent need to migrate every workload on day one. A more practical approach:
- Pick 20 to 50 real prompts from your own work, covering easy, medium, and hard tasks.
- Run them on Luna and Terra first, since the cost savings there are the most immediate and lowest-risk win.
- Escalate only the failures to Sol, and reserve it for the complex work that actually needs frontier-level reasoning.
- Re-check pricing before you commit, because all three tiers were cut within weeks of launch.
Want to see how the wider field compares? Browse the AI tools directory.
Frequently Asked Questions
What's the difference between GPT-5.6 Sol, Terra, and Luna?
Sol is the flagship for the hardest coding, reasoning, and agent work. Terra is the balanced mid-tier for everyday production tasks. Luna is the cheapest, built for high-volume simple tasks where cost per request matters most.
Why was GPT-5.6 delayed before its public release?
The strongest model showed frontier-level capability in vulnerability research, so the U.S. government asked OpenAI to hold broad release until a safety review cleared it. Access was limited to about 20 approved partner organizations until general availability on July 9, 2026.
Is GPT-5.6 available in regular ChatGPT conversations?
Yes. Luna is the default for Free and Go accounts, and Plus and Pro subscribers can use an updated Sol with a slider that sets how much thinking goes into each answer.
Is GPT-5.6 better than Claude for coding?
On coding benchmarks Sol set a new state-of-the-art score at launch, but independent testers were split on day-to-day feel. The best check is to run your own real tasks through both.
Which GPT-5.6 model should I start with?
Start with Luna or Terra on your simpler, repetitive tasks, because the savings are immediate and low-risk, then move only the hard failures up to Sol.
Have GPT-5.6 prices changed since launch?
Yes. Luna was cut 80% and Terra 20% on July 30, and Sol was cut by more than 20% on August 21, so launch-day prices are out of date.
User Reviews
Be the first to review!