GPT-6 is OpenAI's model generation that began with GPT-6 Astra on 3 September 2026, five months after GPT-5.5 and GPT-5.5 Pro were released on 24 April 2026, and it changes prices, reasoning settings, prompt caching parameters and safety classifications for anyone moving off GPT-5.5. GPT-6 is not one model: OpenAI's API documentation lists GPT-6 Astra, GPT-6 Sol, GPT-6.1 Sol, GPT-6 Luna and GPT-6.1 Terra. This page sets GPT-5.5 and GPT-5.5 Pro beside each GPT-6 model with identical fields: API IDs, release dates, context, knowledge cutoff, prices, reasoning effort levels, tools, availability, system card classifications and lifecycle status. It is written for developers planning a migration, buyers estimating cost and architects who need to know which IDs are current. Models are listed by release date, newest first, an objective order that implies no ranking.
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 API documentation as of November 2026. Model details change, so confirm against OpenAI's current documentation.
Key takeaways
- GPT-6 is a family of five API models. Astra, GPT-6 Sol, GPT-6.1 Sol, Luna and GPT-6.1 Terra were released between 3 September and 26 October 2026; GPT-5.5 and GPT-5.5 Pro date from 24 April 2026.
- Context stayed at about 1.05M tokens. GPT-5.5 and every GPT-6 model list a 1,050,000-token context window and 128,000-token maximum output; knowledge cutoffs moved from 1 December 2025 to between 20 April and 1 June 2026.
- Prices now span a 100x range. GPT-5.5 lists $5 input and $30 output per million tokens; GPT-6 ranges from Luna at $0.10/$0.50 to Astra at $10/$50, with GPT-6.1 Sol at $2/$10.
- Some GPT-5.5 settings break on GPT-6. Astra and GPT-6.1 Sol reject reasoning effort none, and OpenAI's migration guide replaces prompt_cache_retention with prompt_cache_options.ttl when moving from GPT-5.5 or earlier.
- Cyber classification rose to Critical. OpenAI's system cards treat GPT-5.5 as High in cybersecurity but below Critical, and treat Astra, GPT-6.1 Sol and GPT-6.1 Terra as Critical.
What is GPT-6 compared with GPT-5.5?
GPT-5.5 is OpenAI's April 2026 flagship, described on its model page as "a new class of intelligence for coding and professional work", with a Pro variant that OpenAI describes as producing "smarter and more precise responses". OpenAI's changelog shows it recommended GPT-5.5 for production API use in chat-latest updates on 5 May, 28 May and 24 June 2026. The GPT-5.6 family (Sol, Terra and Luna, 9 July 2026) sits between the two generations and is covered in the OpenAI Models hub.
GPT-6 splits the line into named tiers. OpenAI's migration guide describes gpt-6-astra for "most demanding work", gpt-6.1-sol for complex work at lower cost, gpt-6.1-terra for visual reasoning and writing, and gpt-6-luna for focused, high-volume tasks. On 7 October 2026 the chat-latest update recommended "the GPT-6 model family". All models here are proprietary: OpenAI publishes no weights and states no parameter count. For a tier-by-tier view inside GPT-6, see GPT-6 Astra vs Sol vs Luna vs Terra.
Every model at a glance
As of 2026-11-15. Source: OpenAI model pages and API changelog.
| Model | API ID | Released | Status on OpenAI pages |
|---|---|---|---|
| GPT-6.1 Terra | gpt-6.1-terra | 2026-10-26 | Current; no deprecation entry |
| GPT-6.1 Sol | gpt-6.1-sol | 2026-09-29 | Current; no deprecation entry |
| GPT-6 Luna | gpt-6-luna | 2026-09-22 | Current; no deprecation entry |
| GPT-6 Sol | gpt-6-sol | 2026-09-22 | Model page points to GPT-6.1 Sol as the newer Sol model; no deprecation entry |
| GPT-6 Astra | gpt-6-astra | 2026-09-03 | Current; no deprecation entry |
| GPT-5.5 Pro | gpt-5.5-pro (snapshot gpt-5.5-pro-2026-04-23) | 2026-04-24 | No deprecation entry |
| GPT-5.5 | gpt-5.5 (snapshot gpt-5.5-2026-04-23) | 2026-04-24 | No deprecation entry |
Every model takes text and image input and returns text; none supports audio or video input on its model page. Open weights: none. Licence: proprietary. Parameters: not stated by OpenAI. GPT-5.5 and GPT-5.5 Pro use dated snapshots; each GPT-6 model page lists a single undated snapshot.
When was each model released?
As of 2026-11-15. Dates as shown in OpenAI's API changelog.
| Date | Event | API ID |
|---|---|---|
| 2026-04-24 | GPT-5.5 and GPT-5.5 Pro released; 1M context, default medium effort, caching only with extended prompt caching | gpt-5.5, gpt-5.5-pro |
| 2026-06-01 | GPT-5.4 and GPT-5.5 available in Amazon Bedrock through an OpenAI-compatible Responses API endpoint | gpt-5.5 |
| 2026-06-24 | chat-latest update recommends GPT-5.5 for production API use | gpt-5.5 |
| 2026-07-30 | Fast mode replaces Priority Processing | All |
| 2026-09-03 | GPT-6 Astra released, with async tool calling, mid-turn steering and mid-conversation effort changes in the Responses API | gpt-6-astra |
| 2026-09-22 | GPT-6 Sol and GPT-6 Luna released | gpt-6-sol, gpt-6-luna |
| 2026-09-25 | Image encoding bug that degraded image understanding in GPT-6 Sol and Luna fixed | gpt-6-sol, gpt-6-luna |
| 2026-09-29 | GPT-6.1 Sol released with Multi-agent beta; Ultrafast mode added for Astra (US data residency only) | gpt-6.1-sol, gpt-6-astra |
| 2026-10-07 | chat-latest update recommends the GPT-6 model family | chat-latest |
| 2026-10-08 | Ultrafast mode added for GPT-6.1 Sol | gpt-6.1-sol |
| 2026-10-26 | GPT-6.1 Terra released | gpt-6.1-terra |
How much does each model cost?
As of 2026-11-15. USD per million text tokens, Standard processing, from each model page; 12-hour cache write and Batch figures from OpenAI's pricing page.
| Model | Input | Cached input | Cache write (30m / 12h) | Output | Batch (input / output) | Notes stated by provider |
|---|---|---|---|---|---|---|
| GPT-6.1 Terra | $1.00 | $0.05 | $1.25 / $2.50 | $5.00 | $0.50 / $2.50 | Fast 2x |
| GPT-6.1 Sol | $2.00 | $0.10 | $2.50 / $5.00 | $10.00 | $1.00 / $5.00 | Fast 2x, Ultrafast 6x |
| GPT-6 Luna | $0.10 | $0.01 | $0.125 / $0.25 | $0.50 | $0.05 / $0.25 | Fast 2x |
| GPT-6 Sol | $2.00 | $0.20 | $2.50 / not stated | $10.00 | 50% of Standard (model page) | Fast 2x; not on the pricing page |
| GPT-6 Astra | $10.00 | $1.00 | $12.50 / $25.00 | $50.00 | $5.00 / $25.00 | Fast 2x; Ultrafast $60 / $300 |
| GPT-5.5 Pro | $30.00 | No cached discount | Not stated by OpenAI on the pages checked (2026-11-15) | $180.00 | Not stated by OpenAI on the pages checked (2026-11-15) | Responses API only |
| GPT-5.5 | $5.00 | $0.50 | Not stated by OpenAI on the pages checked (2026-11-15) | $30.00 | Not stated by OpenAI on the pages checked (2026-11-15) | Not on the pricing page |
Long context: every model page here applies the same rule: prompts over 272K input tokens are billed at 2x input (and cache) rates and 1.5x output for the full request. GPT-6.1 Sol's page adds that subscription usage in Codex and ChatGPT Work carries no long-context surcharge.
Tier labels: OpenAI's pricing page shows tabs named Standard, Flex/Batch, Fast and Ultrafast, but the later tables are headed only "All models"; the Batch column above follows the model pages' statement that Batch and Flex cost 50% of Standard. Regional processing (data residency) endpoints carry a 10% uplift for models released on or after 5 March 2026, which covers every model on this page. See the LLM Batch and Prompt Caching Discounts Compared page.
Worked cost example
At Standard list prices, 10,000 requests a day with a 20K-token prompt and a 1K-token response use 200 million input and 10 million output tokens a day. GPT-5.5 costs $1,000 input plus $300 output, $1,300 a day. GPT-6.1 Sol costs $400 plus $100, $500 a day. GPT-6 Astra costs $2,000 plus $500, $2,500 a day. GPT-6.1 Terra costs $250 a day and GPT-6 Luna $25 a day. Caching changes the result: GPT-6.1 Sol's cached input is 5% of its input rate, against 10% for GPT-5.5. Re-run the arithmetic with token counts measured on your own data.
What changed between models?
Only changes OpenAI documents on its model pages, changelog and migration guide are listed.
- Reasoning effort: GPT-5.5 lists none, low, medium (default), high and xhigh; GPT-5.5 Pro lists medium, high (default) and xhigh. Every GPT-6 model adds max. Astra and GPT-6.1 Sol do not accept none; the migration guide says to use low instead, and to start with low where a request used minimal.
- Sampling parameters: for the GPT-6 models, the migration guide says to remove temperature, top_p and top_logprobs when reasoning effort is not none, and logprobs in Chat Completions.
- Prompt caching: when migrating from GPT-5.5 or earlier, replace prompt_cache_retention with prompt_cache_options.ttl set to "30m", and review cache-write billing. GPT-6 pricing introduces separate 30-minute and 12-hour cache write prices.
- Tool calling: Astra, GPT-6.1 Sol and GPT-6.1 Terra accept Chat Completions requests but need the Responses API for tool calling; GPT-6 Sol and Luna support function calling in Chat Completions only with reasoning effort none.
- New agent controls: async tool calling (async: true, results returned later by call_id), mid-turn steering over a WebSocket connection, and a configuration_update input item that changes reasoning effort mid-conversation. GPT-6.1 Sol adds the Responses Multi-agent beta and reasoning.context set to all_turns.
- Pro tier: GPT-6 has no model named Pro on the pages checked. OpenAI's deprecations page names gpt-6.1-sol with reasoning.mode: pro in the Responses API as the replacement for gpt-5-pro and o3-pro.
For background on effort settings, see Reasoning Models Explained.
What workloads does OpenAI position each model for?
This section records the provider's own positioning wording; it is not an assessment.
| Model | Positioning stated by provider | Related notes stated by provider |
|---|---|---|
| GPT-6.1 Terra | "A fast, efficient model for visual reasoning, precise computer use, and clear, natural writing" | Examples: matching product photos to catalogue entries, interpreting technical drawings, acting on application screens |
| GPT-6.1 Sol | "Near-Astra performance for complex work at a lower cost" | Complex coding, computer use and professional work; 500K default context in Codex and ChatGPT Work |
| GPT-6 Luna | "Our most efficient model for focused, high-volume tasks" | Price rated low on the model page |
| GPT-6 Sol | "Built to power complex coding and agentic workflows" | Page points to GPT-6.1 Sol as the newer Sol model |
| GPT-6 Astra | "Our most capable model for the most demanding work" | Complex reasoning, coding, computer use, research and document creation |
| GPT-5.5 Pro | "Version of GPT-5.5 that produces smarter and more precise responses" | Speed rated Slowest; background mode suggested because requests can take several minutes |
| GPT-5.5 | "A new class of intelligence for coding and professional work" | Recommended for production API use in chat-latest updates, May to June 2026 |
What is the context window of each model?
As of 2026-11-15. From each OpenAI model page.
| Model | Context window | Max output | Knowledge cutoff | Input | Output | Reasoning controls |
|---|---|---|---|---|---|---|
| GPT-6.1 Terra | 1,050,000 tokens | 128,000 tokens | 2026-06-01 | Text, image | Text | none to max, default medium |
| GPT-6.1 Sol | 1,050,000 tokens | 128,000 tokens | 2026-04-30 | Text, image | Text | low to max, default medium |
| GPT-6 Luna | 1,050,000 tokens | 128,000 tokens | 2026-05-18 | Text, image | Text | none to max, default medium |
| GPT-6 Sol | 1,050,000 tokens | 128,000 tokens | 2026-04-20 | Text, image | Text | none to max, default medium |
| GPT-6 Astra | 1,050,000 tokens | 128,000 tokens | 2026-04-30 | Text, image | Text | low to max; default not stated on the model page |
| GPT-5.5 Pro | 1,050,000 tokens | 128,000 tokens | 2025-12-01 | Text, image | Text | medium, high (default), xhigh |
| GPT-5.5 | 1,050,000 tokens | 128,000 tokens | 2025-12-01 | Text, image | Text | none to xhigh, default medium |
The window did not grow between generations; the 272K-token long-context price threshold also applies to all of them. The knowledge cutoff moved forward by between four and six months.
Capabilities and tools compared
As of 2026-11-15. From each model page's features, tools and cloud sections.
| Model | Function calling and structured outputs | Computer use tool | Fine-tuning | Cloud IDs stated by provider |
|---|---|---|---|---|
| GPT-6.1 Terra | Both supported; tool calling via the Responses API | Supported | Not supported | Amazon Bedrock openai.gpt-6.1-terra; Google Cloud gpt-6.1-terra (private preview) |
| GPT-6.1 Sol | Both supported; tool calling via the Responses API | Supported | Not supported | Amazon Bedrock openai.gpt-6.1-sol; Google Cloud gpt-6.1-sol (private preview) |
| GPT-6 Luna | Both supported; Chat Completions function calling only with effort none | Supported | Not supported | Amazon Bedrock openai.gpt-6-luna; Google Cloud gpt-6-luna (private preview) |
| GPT-6 Sol | Both supported; Chat Completions function calling only with effort none | Supported | Not supported | Amazon Bedrock openai.gpt-6-sol |
| GPT-6 Astra | Both supported; tool calling via the Responses API | Supported | Not supported | Amazon Bedrock openai.gpt-6-astra; Google Cloud gpt-6-astra (private preview) |
| GPT-5.5 Pro | Both supported; streaming not supported | Not in the model page's supported tools list | Not supported | Not stated by OpenAI on the pages checked (2026-11-15) |
| GPT-5.5 | Both supported | Supported | Not supported | Amazon Bedrock openai.gpt-5.5 |
GPT-5.5 and every GPT-6 model list the same ten Responses API tools: web search, file search, image generation, code interpreter, hosted shell, apply patch, skills, computer use, MCP and tool search. GPT-5.5 Pro lists six of them (web search, file search, image generation, code interpreter, hosted shell and MCP). For agent risks from these tools, see Computer-Use and Browser Agents and Securing OpenAI Agents SDK and Responses API Guide.
Where is each model available?
| Model | Availability stated by provider |
|---|---|
| GPT-6.1 Terra | OpenAI API (Free tier not supported); Amazon Bedrock; Google Cloud private preview |
| GPT-6.1 Sol | OpenAI API (Free tier not supported); 500K default context in Codex and ChatGPT Work; Amazon Bedrock; Google Cloud private preview |
| GPT-6 Luna | OpenAI API (Free tier not supported); October version replaces GPT-5.6 Luna in ChatGPT (system card); Amazon Bedrock; Google Cloud private preview |
| GPT-6 Sol | OpenAI API (Free tier not supported); October version replaces GPT-5.6 Sol in ChatGPT, September version stays in Codex and ChatGPT Work (system card); Amazon Bedrock |
| GPT-6 Astra | OpenAI API (Free tier not supported); Amazon Bedrock; Google Cloud private preview |
| GPT-5.5 Pro | OpenAI Responses API, including Batch (Free tier not supported) |
| GPT-5.5 | OpenAI API (Free tier not supported); Amazon Bedrock from 2026-06-01 |
Rate limits at the Build tier are 5,000 requests per minute for GPT-5.5 and every GPT-6 model and 500 for GPT-5.5 Pro; GPT-6 Luna lists 2,000,000 tokens per minute at Build against 1,000,000 for the others. Partner clouds publish their own catalogues and dates, so confirm there.
Model cards and system cards
As of 2026-11-15. Classifications are those stated in each OpenAI system card under its Preparedness Framework.
| Model | System card (published) | Cybersecurity | Biological and chemical | AI self-improvement |
|---|---|---|---|---|
| GPT-6.1 Terra | GPT-6.1 Terra (2026-10-26) | Critical | High | Below High |
| GPT-6.1 Sol | GPT-6.1 Sol (2026-09-29) | Critical | High | Below High |
| GPT-6 Luna (October) | GPT-6 October (2026-10-07) | High, below Critical | High | Below High |
| GPT-6 Sol (October) | GPT-6 October (2026-10-07) | High, below Critical | High | Below High |
| GPT-6 Astra | GPT-6 Astra (2026-09-03) | Critical (OpenAI's first model at this level) | High | Below High |
| GPT-5.5 and GPT-5.5 Pro | GPT-5.5 (2026-04-23) | High, below Critical | High | Below High |
The GPT-6.1 Terra card quotes OpenAI's definition of Critical cyber capability as a model that can "identify and develop functional zero-day exploits of all severity levels in many hardened real-world critical systems without human intervention". See the Frontier AI Model Risk Frameworks Guide for how these frameworks work.
Benchmarks
As of 2026-11-15. HealthBench is the one evaluation reported, length-adjusted, in every system card above. Figures are provider-reported; card dates and evaluation configurations differ, so treat cross-row comparisons with care. Independent leaderboards (LMArena, Artificial Analysis, Epoch AI) were not checked for this page.
| Model | HealthBench Professional | HealthBench | HealthBench Hard | Source and date |
|---|---|---|---|---|
| GPT-6.1 Terra | 62.7 | 55.1 | 31.4 | GPT-6.1 Terra system card, 2026-10-26 |
| GPT-6.1 Sol | 64.2 | 58.5 | 36.2 | GPT-6.1 Sol system card, 2026-09-29 |
| GPT-6 Luna (October) | 48.2 | 50.4 | 27.2 | GPT-6 October system card, 2026-10-07 |
| GPT-6 Sol (October) | 55.2 | 52.1 | 27.8 | GPT-6 October system card, 2026-10-07 |
| GPT-6 Astra | 64.7 | 58.3 | 36.6 | GPT-6 Astra system card, 2026-09-03 (values revised 2026-09-22) |
| GPT-5.5 | 51.8 | 56.5 | 31.5 | GPT-5.5 system card, 2026-04-23 |
| GPT-5.5 Pro | Not stated by OpenAI on the pages checked (2026-11-15) | Not stated by OpenAI on the pages checked (2026-11-15) | Not stated by OpenAI on the pages checked (2026-11-15) | Not stated |
Deprecations and retirements
As of 2026-11-15. Source: OpenAI deprecations page. None of the models on this page has a deprecation entry; the table shows the dated entries that name a GPT-6 model as the replacement, which is where most migrations onto GPT-6 come from.
| Model | API ID | Notice or deprecation date | Retirement or shutdown date | Stated replacement |
|---|---|---|---|---|
| GPT-5 | gpt-5 | 2026-06-11 | 2026-12-11 | gpt-6.1-sol |
| GPT-5 Pro | gpt-5-pro | 2026-06-11 | 2026-12-11 | gpt-6.1-sol with reasoning.mode: pro |
| GPT-5 mini | gpt-5-mini | 2026-06-11 | 2026-12-11 | gpt-6.1-terra |
| GPT-5 nano | gpt-5-nano | 2026-06-11 | 2026-12-11 | gpt-6-luna |
| GPT-4o, GPT-4o mini, GPT-4.1, GPT-4.1 mini | Dated snapshots | 2026-10-14 | 2027-01-14 | gpt-6-luna |
| GPT-5.1 and GPT-5.3 Codex | gpt-5.1, gpt-5.3-codex | 2026-10-01 | 2027-04-01 | gpt-6-sol |
| GPT-5.5 Cyber preview | gpt-5.5-cyber-preview | 2026-11-05 | 2027-02-01 | "The most capable cyber model available to you" |
gpt-5.5-cyber-preview is a separate, restricted ID; its deprecation does not apply to gpt-5.5. Track dates on the LLM Deprecation and Retirement Tracker and plan with the AI Model Retirement Guide.
Data handling and security
OpenAI's data controls page states that API data is not used to train its models unless the customer opts in, that abuse monitoring logs are kept for up to 30 days by default, and that Zero Data Retention and Modified Abuse Monitoring require OpenAI's approval. Responses API data is kept for at least 30 days by default or when store is true. The page also states OpenAI may make specific models ineligible for these arrangements for particular customers, with advance written notice, so confirm terms per model. OpenAI's changelog entry for Astra lists misalignment monitoring among its migration notes, and the migration guide notes Fast mode with EU data residency is supported for GPT-6.1 Sol, Terra and Luna but not for Astra or GPT-6 Sol.
Switching models does not change who holds the API key that calls them; key handling is the control that carries across every version. See the OpenAI API Keys Security Guide, How to audit OpenAI and Anthropic organisation API keys and the LLMjacking Guide.
How to choose a model: factual criteria
| Criterion | What the documentation shows |
|---|---|
| Code still sends reasoning effort none | Accepted by GPT-5.5, GPT-6 Sol, Luna and GPT-6.1 Terra; rejected by Astra and GPT-6.1 Sol |
| Code sets temperature or top_p | Remove for GPT-6 models when effort is not none |
| Code uses prompt_cache_retention | Replace with prompt_cache_options.ttl when moving from GPT-5.5 |
| Function calling through Chat Completions | GPT-6 Sol and Luna only with effort none; others need the Responses API |
| List price per token | From $0.10/$0.50 (Luna) to $10/$50 (Astra); GPT-5.5 $5/$30; GPT-5.5 Pro $30/$180 |
| Knowledge cutoff | 2025-12-01 for GPT-5.5; 2026-04-20 to 2026-06-01 for GPT-6 |
| Low-latency tiers | Ultrafast priced for Astra and GPT-6.1 Sol only |
| Deprecation replacement mapping | OpenAI names gpt-6.1-sol, gpt-6.1-terra, gpt-6-luna or gpt-6-sol depending on the retiring model |
Run your own evaluation on representative prompts at the same reasoning effort before switching; system card evaluations are not a substitute for workload testing.
Practitioner checklist
- Inventory every service and agent calling gpt-5.5, gpt-5.5-pro or a model with a GPT-6 replacement and record the pinned ID in the AI register.
- Search code for reasoning effort none, minimal, temperature, top_p, logprobs and prompt_cache_retention before switching to a GPT-6 model.
- Move tool-calling workloads to the Responses API where the target is Astra, GPT-6.1 Sol or GPT-6.1 Terra.
- Re-cost workloads with the 272K long-context multiplier, 30-minute and 12-hour cache writes and the 10% regional processing uplift.
- Test on the target model at matched reasoning effort, and keep the GPT-5.5 path available for rollback until results are signed off.
- Confirm data retention and data residency terms per model, including Fast and Ultrafast residency limits.
- Store OpenAI keys in a secrets manager, scope them per project and set spend limits.
Standards and references
- OpenAI: Models (accessed 2026-11-15)
- OpenAI: GPT-5.5 model page (accessed 2026-11-15)
- OpenAI: GPT-5.5 Pro model page (accessed 2026-11-15)
- OpenAI: GPT-6 Astra model page (accessed 2026-11-15)
- OpenAI: GPT-6.1 Sol model page (accessed 2026-11-15)
- OpenAI: GPT-6 Sol model page (accessed 2026-11-15)
- OpenAI: GPT-6 Luna model page (accessed 2026-11-15)
- OpenAI: GPT-6.1 Terra model page (accessed 2026-11-15)
- OpenAI: Using the latest model (accessed 2026-11-15)
- OpenAI: Pricing (accessed 2026-11-15)
- OpenAI: API changelog (accessed 2026-11-15)
- OpenAI: Deprecations (accessed 2026-11-15)
- OpenAI: Data controls in the OpenAI platform (accessed 2026-11-15)
- OpenAI: GPT-5.5 system card (accessed 2026-11-15)
- OpenAI: GPT-6 Astra system card (accessed 2026-11-15)
Related NHI Mgmt Group resources: OpenAI Models, GPT-6 Astra vs Sol vs Luna vs Terra, GPT-6 Astra, Frontier AI Models Compared, LLM Pricing Tracker.