GPT-6 Luna: Specs, Pricing, Limits and Safety Card Facts

GPT-6 Luna is a proprietary large language model from OpenAI, released through the OpenAI API on 22 September 2026, that OpenAI describes as "our most efficient model for focused, high-volume tasks". This profile collects the published facts on GPT-6 Luna in one place: API identifier, context window, output limit, knowledge cutoff, prices, rate limits, tools, availability, system card results and deprecation status. It is written for developers, architects, buyers and governance teams who need exact figures rather than summaries. Every fact comes from OpenAI's own model page, pricing page, changelog, deprecations page, data controls page and Deployment Safety Hub, listed under Standards and references. Where OpenAI publishes nothing on a point, the page says so.

Last checked: 2026-11-15.

This is an independent reference page from NHI Mgmt Group based on public provider documentation; it is not sponsored or endorsed by any provider. It uses OpenAI's public GPT-6 Luna model page as of November 2026. Product details change, so confirm against OpenAI's current documentation.

Key takeaways

• Released 22 September 2026. OpenAI's API changelog records the release of GPT-6 Sol and GPT-6 Luna together on 22 September 2026.
• 1,050,000-token context, 128,000-token output. The model page lists a 1.05M context window, 128K max output and a knowledge cutoff of 18 May 2026.
• $0.10 input and $0.50 output per million tokens. Cached input is $0.01, Batch and Flex are 50% of Standard, Fast mode is 2x, and prompts over 272K input tokens cost more for the whole request.
• Rated High for cybersecurity and biology. OpenAI's 29 October 2026 addendum classifies GPT-6 Luna as High in cybersecurity and in biological and chemical, and below High in AI self-improvement.
• Named replacement for six retiring model lines. OpenAI's deprecations page lists gpt-6-luna as the replacement for GPT-4o, GPT-4o mini, GPT-4.1, GPT-4.1 mini, gpt-5-nano and gpt-5.4-nano snapshots; Luna itself has no shutdown date.

What is GPT-6 Luna?

GPT-6 Luna is one of four flagship models on OpenAI's models page, alongside GPT-6 Astra, GPT-6.1 Sol and GPT-6.1 Terra, and it carries the lowest listed price of the four. The models page points cost-sensitive, high-volume workloads to GPT-6 Luna. OpenAI rates its reasoning as "High" and its speed as "Fast" on the model page.

In ChatGPT, OpenAI's 7 October 2026 system card update states that GPT-6 launched to everyone across free and paid plans, replacing GPT-5.6 Sol and GPT-5.6 Luna there. A further update on 29 October 2026 "provides more concise responses and reduces overly agreeable language". GPT-6 Luna is also the model behind OpenAI's Decisions API. For the whole OpenAI line-up see OpenAI Models, and for a side-by-side view of the four tiers see GPT-6 Astra vs Sol vs Luna vs Terra.

Key facts at a glance

As of 15 November 2026.

Release and version history

GPT-6 Luna follows GPT-5.6 Luna in OpenAI's naming. The API changelog and Deployment Safety Hub show this sequence.

The 7 October card notes that Codex and ChatGPT Work still use the September versions. OpenAI does not state on the pages checked (2026-11-15) whether the API snapshot gpt-6-luna tracks the ChatGPT updates.

How much does GPT-6 Luna cost?

Prices in US dollars per million text tokens, as of 15 November 2026, from the OpenAI pricing page and the model page.

• Long context: the model page states that prompts over 272K input tokens are charged at 2x the input and cache rates and 1.5x the output rate for the full request.
• Regional processing: a 10% premium applies where regional processing is available.
• Priority: the pricing page states that Priority processing was renamed Fast mode on 30 July 2026.
• Tools: built-in tool calls are priced separately, for example web search at $10.00 per 1,000 calls plus search content tokens at model rates, and file search at $2.50 per 1,000 calls. At Luna's token prices, tool fees can be a large share of the bill.

For how caching and batch discounts work across providers, see Model Pricing Explained and the LLM Pricing Tracker.

Context window and output limits

GPT-6 Luna accepts up to 1,050,000 tokens of context and returns up to 128,000 output tokens, and it supports reasoning tokens. The 272K input threshold is a pricing boundary, not a capability limit. For background, see Context Windows Explained and Knowledge Cutoff Explained.

Rate limits on the model page, as of 15 November 2026:

OpenAI states that usage tiers rise automatically as requests and spend increase.

Modalities and tools

• Modalities: text input and output, image input. Audio and video are not listed as supported.
• Features: streaming, function calling and structured outputs are supported; fine-tuning is not offered for this model.
• API surface: built-in tools and function calling require the Responses API; Chat Completions supports function calling only when reasoning_effort is none.
• Built-in tools (Responses API): web search, file search, image generation, code interpreter, hosted shell, apply patch, skills, computer use, MCP and tool search.
• Reasoning effort: six values from none to max, with medium as the default. When reasoning effort is not none, the "Using GPT-6" guide says to remove temperature, top_p and top_logprobs.
• Decisions API: generally available with gpt-6-luna since 22 October 2026, per the changelog.

Where is GPT-6 Luna available?

Cloud catalogues can lag or differ by region, so confirm availability in the cloud provider's own console before committing.

Access restrictions

GPT-6 Luna is a general availability API model. The model page shows it as unavailable on the Free API usage tier and available from the Build tier upwards. No trusted-access or verification programme is stated for it. EU data residency is available with Standard, Fast, Flex and Batch processing, per the model page. Zero data retention and modified abuse monitoring need prior approval from OpenAI (see Data handling).

Benchmarks (provider-reported, dated)

OpenAI's model page and "Using GPT-6" guide give no capability benchmark table for GPT-6 Luna. The figures below are safety and preparedness evaluations from OpenAI's Deployment Safety Hub, for the ChatGPT versions dated 7 October and 29 October 2026. They measure risk thresholds, not general task quality. Independent leaderboard results are not included on this page. For how to read such numbers, see How to Read LLM Benchmarks.

Model card and system card

OpenAI documents GPT-6 Luna on its Deployment Safety Hub in two entries checked for this page: "GPT-6 Sol and GPT-6 Luna: October 2026 update" (7 October 2026) and "Addendum to GPT-6 Luna System Card: October 29, 2026 Update" (29 October 2026). What they state:

• Preparedness classifications: High in cybersecurity and in biological and chemical, below High in AI self-improvement; the 7 October card states Luna is below Critical in cybersecurity.
• Safeguards: the 7 October card says the same safeguards as the GPT-5.6 System Card apply; the 29 October addendum keeps the 7 October safeguards.
• Critical biology testing: the 7 October card states separate Critical testing was not required because Luna scored below Sol on all High biology evaluations; the 29 October addendum reports Critical-level results below threshold.
• Regressions reported (7 October): statistically significant regressions on self-harm, gore and sexual content, and on several under-18 categories, which the card calls low severity.
• Changes reported (29 October): improvements in under-18 safe completions and prompt-injection robustness; regressions in dual-use biology refusals and in factual error rates, which the addendum says remain lower than GPT-5.6 Luna.
• Caveat: the 29 October addendum notes the multiturn jailbreak attacker run was cut short by infrastructure issues.

See Frontier AI Model Risk Frameworks Guide for how preparedness frameworks work.

Data handling

• Training use: OpenAI states that data sent to the API is not used to train or improve its models by default (since 1 March 2023) unless the customer opts in.
• Abuse monitoring: logs are retained for up to 30 days by default, longer where required by law or to protect the services.
• Zero data retention: zero data retention and modified abuse monitoring require prior approval from OpenAI. With zero data retention, the Responses API treats store as false.
• Application state: Responses API state is kept for at least 30 days by default; prompt caching on Chat Completions may keep encrypted tensors on GPU-local storage for up to 24 hours.
• Data residency: EU data residency is available with Standard, Fast, Flex and Batch processing; regional processing carries a 10% premium.

Model-specific retention terms beyond these platform-wide rules are not stated by OpenAI on the pages checked (2026-11-15). High-volume workloads multiply exposure if a key leaks; see the OpenAI API Keys Security Guide and AI Data Retention and Deletion.

Deprecation status

GPT-6 Luna has no deprecation notice or shutdown date on OpenAI's deprecations page as of 15 November 2026. It is listed as the replacement for these models:

GPT-5.6 Luna has no deprecation entry on the same page as of the check date.

How GPT-6 Luna differs from other models in its family

Figures from OpenAI's model pages and pricing page, as of 15 November 2026, listed by standard input price. Descriptions are OpenAI's own wording.

Its rate limits are higher than GPT-6.1 Terra's at every paid tier listed on the two model pages. See the GPT-6.1 Terra, GPT-6.1 Sol and GPT-6 Astra profiles for each model's details, and OpenAI Mini and Nano Models Compared for the retiring small models Luna replaces.

Practitioner checklist

• Pin the API ID gpt-6-luna and record the check date of every limit and price you rely on.
• Plan migrations for gpt-5-nano (11 December 2026), GPT-4o, GPT-4o mini, GPT-4.1 and GPT-4.1 mini (14 January 2027) and gpt-5.4-nano (1 April 2027), testing each on gpt-6-luna.
• Use the Responses API for built-in tools, or set reasoning_effort to none if you need function calling on Chat Completions.
• Budget tool call fees and long-context rates separately, since they can outweigh Luna's token cost.
• Do not assume API behaviour matches ChatGPT: the system card updates describe ChatGPT versions.
• Enable EU data residency where required and account for the 10% regional premium.
• Apply for zero data retention before sending regulated data, and check which endpoints and tools it covers.
• Store API keys in a secrets manager and set spend limits. See How to set spend limits on LLM API keys.

Standards and references

• OpenAI: GPT-6 Luna model page (accessed 2026-11-15)
• OpenAI Deployment Safety Hub: GPT-6 Sol and GPT-6 Luna: October 2026 update (accessed 2026-11-15)
• OpenAI Deployment Safety Hub: Addendum to GPT-6 Luna System Card: October 29, 2026 Update (accessed 2026-11-15)
• OpenAI: Pricing (accessed 2026-11-15)
• OpenAI: API changelog (accessed 2026-11-15)
• OpenAI: Deprecations (accessed 2026-11-15)
• OpenAI: Models (accessed 2026-11-15)
• OpenAI: Data controls in the OpenAI platform (accessed 2026-11-15)

Related NHI Mgmt Group resources: OpenAI Models, GPT-6 Astra vs Sol vs Luna vs Terra, Reasoning Models Explained, LLMjacking Guide.