GPT-5.6 in Codex: Credit Pricing, Multi-Agent, and Weekly Limit Facts

GPT-5.6 is available in Codex across Sol, Terra, and Luna, but Codex subscription usage is not the same as direct API billing. Codex converts model token usage into credits, draws from included plan usage first, and can use purchased credits after the included allowance is exhausted on eligible plans.

The most important limit clarification is simple: OpenAI has not announced that weekly limits were removed for every Codex user. The current Codex pricing page still describes a five-hour usage window and says additional weekly limits may apply. A separate Enterprise/Edu admin change is replacing old workspace-configured weekly credit controls with monthly controls; that is not the same as making personal Codex usage unlimited.

GPT-5.6 Codex Credit Rates

The current token-based rate card lists credits per 1 million tokens:

Model	Input	Cached input	Output
GPT-5.6 Sol	125 credits	12.5 credits	750 credits
GPT-5.6 Terra	62.5 credits	6.25 credits	375 credits
GPT-5.6 Luna	25 credits	2.5 credits	150 credits

Sol has the same Codex credit rate as GPT-5.5. Terra costs half as many credits per token, and Luna costs one fifth as many. Output-heavy tasks consume the allowance fastest.

What Happened to the Weekly Limit?

There are three different systems that are easy to confuse:

1. Codex Included Usage

ChatGPT plans include a plan-dependent Codex allowance. Usage is measured over a rolling five-hour window, and OpenAI says additional weekly limits may apply. The exact remaining amount is shown in Codex Settings > Usage because it depends on the plan, model, workload, and current product policy.

2. Purchased Credits

Eligible Plus and Pro users can buy credits after consuming included Codex usage. This removes the need to stop working or upgrade immediately, but it is paid overage, not unlimited included usage. Auto top-up may also be available.

3. Enterprise/Edu Admin Controls

OpenAI is migrating the old Permissions & roles weekly credit-control setting to monthly Usage limits for Enterprise and Edu workspaces. When that migration completes, the old admin-configured weekly setting no longer applies. Model allowances, product rate limits, workspace budgets, and spend controls can still apply.

Therefore, it is accurate to say one legacy weekly admin control is being replaced. It is not accurate to say all Codex weekly limits are gone.

GPT-5.6 Features in Codex

• Sol, Terra, and Luna are available to eligible paid plans; availability can vary by plan and rollout.
• max reasoning is available to users with GPT-5.6 access in Codex.
• ultra is available in Codex on Plus and higher plans and coordinates parallel agents for demanding work.
• Fast mode consumes credits at a higher rate on supported models.
• Code review currently uses GPT-5.3-Codex according to the official rate card, even when interactive Codex work uses GPT-5.6.

Codex Multi-Agent vs API Multi-Agent

Codex ultra and the Responses API multi-agent beta are related capabilities but different billing surfaces.

In Codex, usage is converted into subscription credits. In the API, developers pay the GPT-5.6 token rates and build concurrent subagent workflows through the Responses API beta. Parallel agents can finish difficult work faster, but they also create multiple token-consuming workstreams. Budget for total tokens across all subagents, not only the final synthesized answer.

Which Model Preserves Codex Usage Best?

Use Luna for repository search, file summaries, classification, and routine edits. Use Terra as the default for normal implementation and debugging. Reserve Sol for tasks where higher capability materially reduces retries or failure cost.

This routing policy can stretch the same included allowance much further than running every task on Sol. It also makes purchased-credit spending more predictable.

For direct API costs, see the GPT-5.6 API Pricing Guide and Sol vs Terra vs Luna comparison.

Official Sources

• Codex pricing and usage limits
• OpenAI Codex rate card
• Using credits for flexible Plus/Pro usage
• Enterprise and Edu usage-limit management
• GPT-5.6 announcement