OpenClaw + GPT-5.6: Run OpenAI's Sol, Terra & Luna Models

What Is GPT-5.6?

GPT-5.6 is OpenAI's current model release, and it landed in OpenClaw with upstream version v2026.7.1 (released 2026-07-14). It ships as three distinct models rather than one:

• GPT-5.6 Sol — the flagship. OpenAI describes it as its "latest frontier agentic coding model."
• GPT-5.6 Terra — a balanced mid-tier model for everyday agent work.
• GPT-5.6 Luna — the fastest, most efficient of the three.

In the OpenClaw Launch dashboard model picker, these appear as openai-codex/gpt-5.6-sol, openai-codex/gpt-5.6-terra, and openai-codex/gpt-5.6-luna.

Is GPT-5.6 Free? (Updated 6 August 2026)

Partly. On 6 August 2026 OpenAI made GPT-5.6 Luna the default model for ChatGPT Free and Go users, with unlimited text chats and a new Think button that lets the free tier request higher reasoning effort. That is a real change, but it is narrower than the headlines suggest, and it is worth being precise about because the two halves point in opposite directions:

• In the ChatGPT app — effectively free. Luna is unlimited for text chat on Free and Go. OpenAI notes that limits still apply to file uploads, image generation, and other tools, and that the unlimited text is subject to abuse guardrails. Only Luna is included; Terra and Sol are not.
• In the API — still paid. There is no free API tier and no free variant of Luna. What did change is the price: on 30 July 2026 OpenAI cut Luna by roughly 80% and Terra by 20%.

Current GPT-5.6 API pricing

Cached reads are 90% cheaper than fresh input, which matters for an agent that re-sends a long system prompt on every turn. Fast mode is billed at double the standard rate.

Two Ways to Run GPT-5.6

There are two supported paths, and they have very different cost profiles:

1. ChatGPT subscription (any Codex-eligible plan: Go, Plus, Pro, Business, or Enterprise/Edu — not Free) — connect your existing ChatGPT plan to the bot and GPT-5.6 costs $0 per token. Usage draws from your plan's weekly allowance instead of a metered API bill. In the model picker, subscription-backed models are grouped under "OpenAI · Subscription." See the full walkthrough in Connect a ChatGPT Subscription to OpenClaw.
2. OpenAI API key — bring your own key and pay OpenAI directly, billed per token at the rates above. This is the standard BYOK path if you'd rather not tie usage to your ChatGPT plan. After the 30 July price cut, Luna at $0.20 in / $1.20 out is cheap enough that a light personal agent often costs less per month here than a subscription would.

The subscription route is also the one with the largest context window: GPT-5.6 gets a 272k token context window when running through the ChatGPT subscription (Codex) route.

Deploy GPT-5.6 From the Dashboard

1. Open your instance card at openclawlaunch.com/dashboard.
2. Connect a ChatGPT subscription (or add an OpenAI API key) from the instance card.
3. Open the model picker and choose GPT-5.6 Sol, Terra, or Luna — subscription-backed options show under "OpenAI · Subscription."
4. The switch applies immediately, no redeploy required.

Reasoning Effort: the /think Command

GPT-5.6's reasoning effort is adjustable per conversation with the in-chat /think command. Levels are low, medium, high, xhigh, and max — plus an ultra level available on Sol and Terra. OpenAI's own guidance is that Sol is highly capable even at lower reasoning efforts, so the practical move is to start low and turn it up only for harder jobs.

OpenClaw Launch sets a medium default reasoning level on instances upgraded to v2026.7.1, which balances speed and depth for typical agent work. You can override it two ways:

• Per chat — type /think high (or any level) in the conversation.
• Permanently — use the Files button on the instance card to open openclaw.json and edit agents.defaults.thinkingDefault.

Which Model Should You Use?

Frequently Asked Questions

Does OpenClaw Launch support GPT-5.6 today?

Yes. GPT-5.6 shipped in upstream OpenClaw v2026.7.1 (2026-07-14), and OpenClaw Launch instances running that version or newer can select it in the model picker — no waitlist, no manual config edits required to enable it.

Is GPT-5.6 free now?

In the ChatGPT app, GPT-5.6 Luna is free and unlimited for text chat on the Free and Go tiers as of 6 August 2026, with limits still applying to uploads, images, and other tools. In the API it is not free — Luna is $0.20 per 1M input tokens and $1.20 per 1M output tokens. Terra and Sol are not part of the free tier at all.

Can I use the free ChatGPT plan to power an OpenClaw agent?

No. Connecting a ChatGPT plan to a bot goes through OpenAI's Codex sign-in, and the Free plan is not eligible for it — OpenAI lists Go, Plus, Pro, Business, and Enterprise/Edu. Free unlimited Luna is a chatgpt.com benefit only. To run GPT-5.6 on an agent you need either a Codex-eligible subscription or an OpenAI API key.

Does GPT-5.6 cost extra on top of my plan?

Not if you connect a Codex-eligible ChatGPT plan (Go, Plus, Pro, Business, or Enterprise/Edu) — usage draws from your plan's allowance at $0 per token. If you use an OpenAI API key instead, you're billed by OpenAI per token as usual.

What's the difference between Sol, Terra, and Luna?

Sol is the flagship, built for harder agentic coding and reasoning work. Terra is a balanced everyday option. Luna is the fastest and most efficient, aimed at simpler, higher-volume tasks. All three are switchable from the same dashboard model picker.

How do I change the reasoning effort?

Use /think plus a level (low, medium, high, xhigh, max, or ultra on Sol and Terra) in any chat to change it for that conversation. To change the permanent default, edit agents.defaults.thinkingDefault in openclaw.json via the instance card's Files button.

What's Next?

• Connect a ChatGPT subscription — full walkthrough for the $0-per-token route
• OpenClaw + OpenAI guide — full setup for any OpenAI model
• Compare all models — pricing, context, and strength side by side
• See pricing — deploy with AI credits included, starting at $3/month