GPT-6 Luna is rolling out in Codex and ChatGPT Work. Paid plans get access subject to rollout and workspace settings; Free and Go access is desktop only at Standard speed. GPT-5.6 Luna remains available during rollout. The estimates below distinguish the generations.
| Model | Plus | Pro $100 | Previous Pro 20x estimates | Business Standard |
|---|---|---|---|---|
GPT-6 Lunagpt-6-luna | 350–3,000 | 1,750–14,000 | 7,000–56,000 | 350–3,000 |
GPT-5.6 Lunagpt-5.6-luna | 250–2,000 | 1,250–10,000 | 5,000–40,000 | 250–2,000 |
OpenAI estimates of local messages per five-hour period, not fixed caps. Actual usage varies; weekly limits may also apply. Pro $100, $200, $500 and Business Premium have no five-hour usage limit. The Pro columns preserve earlier published estimates, not current caps. Pro plans still have included usage allowances; check Settings > Usage for remaining allowance and reset times. Oldest verification in these tables: 2026-09-27. OpenAI usage guidance.
Pro $200 is available for new subscriptions with a lower allowance for non-grandfathered accounts; OpenAI has not published its replacement multiplier. Eligible grandfathered subscribers keep the previous allowance through October 29, 2026 while subscribed, then move to the lower allowance. Earlier published Pro 5x and Pro 20x message estimates are historical evidence, not current five-hour caps or proof of current Pro $200 entitlement. Pro $100, $200, and $500 currently have no five-hour usage limit in Work and Codex, but still have included usage allowances. Pro $500 has the highest included Pro usage and Astra Ultrafast, with no published multiplier or model-specific message estimate. OpenAI Pro tiers.
Free Luna access: desktop only, Standard speed; subject to rollout. No numeric estimate is published.
Go Luna access: desktop only, Standard speed; subject to rollout. No numeric estimate is published.
| Standard speed, per 1M tokens | Input | Cached input | Output |
|---|---|---|---|
| GPT-6 Luna | 2.5 credits | 0.25 credits | 12.5 credits |
| GPT-5.6 Luna | 5 credits | 0.5 credits | 30 credits |
ChatGPT Work and Codex credit rates, not API dollars or a formula for included allowance. How credits work.
OpenAI's Reserve article still names GPT-5.6 Luna. Regular GPT-6 Luna access does not prove that Luna Reserve has migrated. Read the Sol and Luna rollout details or the changes ledger.
Earlier GPT-5.6 Luna estimates and measurements
GPT-5.6 Luna Codex usage draws from the same shared Work and Codex plan allowance as Terra, Sol, and Astra — OpenAI publishes estimated 5-hour local-message ranges of 250 to 2,000 on Plus and Business Standard, up to 5,000 to 40,000 on Pro $200 (previous 20x allowance). GPT-6 Luna has lower published credit rates; the older model remains available during rollout. These are OpenAI's estimates, not fixed message quotas — actual usage varies with the task, and weekly limits may also apply.
| Plan Tier | Luna 5-Hour Estimate |
|---|---|
| ChatGPT Plus ($20/mo) | 250 to 2,000 local messages |
| ChatGPT Pro $100 (5x) | 1,250 to 10,000 local messages |
| ChatGPT Pro $200 (previous 20x allowance) | 5,000 to 40,000 local messages |
| Business Standard | 250 to 2,000 local messages |
| Business Premium | No 5-hour limit (5x Standard included usage; weekly cap not published) |
Weekly allowances are not published by OpenAI for any tier. Run /status in the CLI for your account's actual remaining percentages and reset times.
Does GPT-5.6 Luna Have Its Own Quota?
No. Standard Luna usage is not a separate quota — it draws from the same shared Work and Codex plan allowance that Terra, Sol, and Astra all draw from. Switching your active model to Luna does not unlock additional capacity; it changes how much of your existing allowance each message consumes.
The one genuinely separate Luna-specific allowance is Luna Reserve, a fallback capacity available only to select personal Plus and Pro accounts. Luna Reserve provides selected accounts additional Luna usage after regular usage is exhausted — it is not available in Business or Enterprise workspaces, and it has its own usage limit separate from the shared allowance covered on this page.
One Controlled Test: Luna's Measured 5-Hour Impact
In a controlled ChatGPT Plus experiment on one frozen coding task, GPT-5.6 Luna passed all 112 deterministic tests at Medium reasoning effort — the same result Sol and Astra achieved in the same run. Luna's run did not move the displayed 5-hour percentage at Codex Usage's 1-percentage-point resolution, while it took the longest to finish: about 28% slower than Sol and 8% slower than Astra.
This is one controlled run on one task, not a universal “Luna uses zero quota” claim. See the full first-party measurement across Luna, Sol, and Astra for the complete table and methodology.
GPT-5.6 Luna vs. Terra, Sol, and Astra
GPT-5.6 Luna has higher published message estimates than GPT-5.6 Terra, GPT-5.6 Sol and GPT-6 Astra. GPT-6 Luna has higher estimates still, as shown above. This table compares the earlier generation with Astra.
| Plan Tier | GPT-5.6 Luna | GPT-5.6 Terra | GPT-5.6 Sol | GPT-6 Astra |
|---|---|---|---|---|
| ChatGPT Plus ($20/mo) | 250 to 2,000 | 25 to 200 | 10 to 100 | 5 to 45 |
| ChatGPT Pro $100 (5x) | 1,250 to 10,000 | 125 to 1,000 | 50 to 500 | 25 to 225 |
| ChatGPT Pro $200 (previous 20x allowance) | 5,000 to 40,000 | 500 to 4,000 | 200 to 2,000 | 100 to 900 |
| Business Standard | 250 to 2,000 | 25 to 200 | 10 to 100 | 5 to 45 |
Business Premium removes the 5-hour limit entirely for all models (5x Standard included usage, exact weekly cap not published). See ChatGPT Business Codex limits for details.
Luna Credit Rates Across Generations
OpenAI publishes substantially higher estimated local-message ranges for Luna than for Terra, Sol, or Astra, and GPT-5.6 Luna has lower credit rates than the older Sol and Terra models. GPT-6 Luna has lower rates still. These figures describe published estimates and rates, but they are not a fixed conversion formula for included subscription allowance.
| Current Work & Codex Rate (per 1M tokens) | Input | Cached Input | Output |
|---|---|---|---|
| GPT-6 Luna | 2.5 credits | 0.25 credits | 12.5 credits |
| GPT-5.6 Luna | 5 credits | 0.5 credits | 30 credits |
| GPT-5.6 Terra | 50 credits | 5 credits | 300 credits |
| GPT-5.6 Sol | 100 credits | 10 credits | 500 credits |
| GPT-6 Astra | 250 credits | 25 credits | 1,250 credits |
Compare the published input, cached-input and output credit rates above. Both GPT-5.6 models remain available during rollout; these are rates for the earlier generation.
GPT-5.6 Luna has lower published credit rates than Astra. GPT-6 Luna has lower rates still. Lower token rates do not establish a fixed included-usage multiplier.
These figures describe metered credit rates, not a compute-footprint explanation or a guaranteed multiplier on included allowance. Purchased credits and eligible token-based Work and Codex activity use the rate card above; direct API usage has separate billing and rate limits. See Codex credits, paid instant resets, banked resets, and API billing for how these fit together.
When to Use Luna Instead of Terra, Sol, or Astra
OpenAI describes Luna as its fastest and lowest-cost GPT-5.6 model, optimized for cost-sensitive, high-volume workloads. Editorially, that fits clearly-scoped, repeatable tasks where you already know what a good result looks like — formatting, extraction, test scaffolding, routine repository discovery, and other repetitive work are illustrative examples, not an official OpenAI task list. For ambiguous, high-value, or reasoning-heavy work, OpenAI's own guidance points toward Sol or Astra instead. See the full Codex recommended model and Power presets guide for OpenAI's model-by-model positioning and how to select Luna explicitly.
What OpenAI Has Not Published About Luna
How to Check Your Remaining Codex Quota
Because Luna draws from your overall Work and Codex allowance, check your live 5-hour and weekly percentages by running the status command in the terminal:
/statusDisplays your current 5-hour and weekly remaining percentages and reset times. Run it in your own CLI session for your actual numbers.
You can also check the official ChatGPT usage settings page or use our free local Codex Usage desktop monitoring tool. Neither OpenAI nor the local tool attributes usage specifically to Luna on a per-prompt or token level.
Frequently Asked Questions
What is GPT-5.6 Luna in Codex?
GPT-5.6 Luna is OpenAI's fastest and lowest-cost GPT-5.6 model, optimized for cost-sensitive, high-volume workloads. In Codex and ChatGPT Work, it remains available during the GPT-6 Sol and GPT-6 Luna rollout alongside other models, all drawing from the same shared Work and Codex plan allowance.
How many messages do I get with GPT-6 Luna or GPT-5.6 Luna?
For GPT-6 Luna, OpenAI estimates 350 to 3,000 local messages per five-hour period on Plus and Business Standard, 1,750 to 14,000 in the earlier Pro 5x reference, and 7,000 to 56,000 in the previous Pro 20x reference. Pro $100, $200, and $500 currently have no five-hour usage limit in Work and Codex, but still have included allowances; check Settings > Usage for remaining allowance and reset times. For GPT-5.6 Luna, those estimates are 250 to 2,000, 1,250 to 10,000, and 5,000 to 40,000, respectively. These are estimates, not fixed quotas; actual usage varies and weekly limits may also apply.
Does Luna have its own separate quota?
No. Standard Luna usage draws from the same shared Work and Codex plan allowance as Terra, Sol, and Astra — it is not a separate quota. The one Luna-specific allowance that is separate is Luna Reserve, a fallback capacity available to select personal Plus and Pro accounts after regular usage is exhausted. See the Luna Reserve guide for that distinct feature.
How do Luna's credit rates compare across generations?
GPT-6 Luna costs 2.5 credits per million input tokens and 12.5 credits per million output tokens at Standard speed. GPT-5.6 Luna costs 5 credits and 30 credits, respectively. Both have lower published rates than Sol and Astra. These credit rates are not API dollars or a formula for included allowance. See the Codex credits and billing guide for how included usage and metered billing differ.
What kinds of tasks is Luna best for?
OpenAI positions Luna for cost-sensitive, high-volume workloads. Editorially, that tends to fit clearly-scoped, repeatable tasks such as formatting, extraction, test scaffolding, routine repository discovery, and other repetitive work where you already know what a good result looks like — these are illustrative examples, not an official OpenAI task list. For ambiguous or high-stakes work, OpenAI's own guidance points toward Sol or Astra instead.
How is Luna different from Luna Reserve?
Regular GPT-6 Luna and GPT-5.6 Luna usage draws from your shared Work and Codex allowance. Luna Reserve is a separate fallback allowance for selected personal Plus and Pro accounts after regular usage is exhausted. OpenAI's Reserve article still names GPT-5.6 Luna; the rollout does not confirm a Reserve migration. See the dedicated Reserve guide for eligibility and troubleshooting.
How do I select Luna instead of the default model?
Choose gpt-6-luna in the model picker when available, use /model in Codex CLI, pass --model gpt-6-luna at launch, or set model = "gpt-6-luna" in config.toml. GPT-5.6 Luna remains available during rollout; use its explicit ID if you want the earlier model. See the recommended model guide for selection methods and access qualifications.
- Codex usage limits table — every plan in one table.
- ChatGPT Plus Codex limits — the full Plus breakdown.
- ChatGPT Pro Codex limits guide — current Pro allowances and the legacy Pro 20x estimates.
- ChatGPT Business Codex limits — Standard vs. Premium seats.
- GPT-6 Astra Codex usage limits — including Astra vs. Sol usage economics.
- Codex Luna Reserve (gpt-reserve) — the separate fallback allowance, not covered on this page.
- Codex recommended model and Power presets — how to select Luna explicitly.
- when your Codex usage resets — recovery timing for both windows.
- Codex credits, paid instant resets, banked resets, and API billing — included usage vs. metered token billing.
- OpenAI Codex Models Reference — OpenAI's own model documentation.