OpenAI Models: Every Model, Spec, Price and Release Date

OpenAI is an AI company whose GPT large language models are offered through the OpenAI API, ChatGPT and cloud platforms, and its model line-up changed several times in 2026. This hub lists every OpenAI text and reasoning model in OpenAI's public documentation as of 15 November 2026, with API IDs, release dates, context windows, prices, availability, access restrictions and shutdown dates. It is written for developers, buyers, architects and anyone who needs to know which GPT model is which. Models are listed by release date, newest first, an objective order that implies no ranking. Every fact comes from OpenAI's own documentation, cited in the sources list.

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 documentation as of November 2026. Model details change, so confirm against OpenAI's current documentation.

Key takeaways

• Four flagship models are current. OpenAI's models page lists GPT-6 Astra, GPT-6.1 Sol, GPT-6.1 Terra and GPT-6 Luna as flagship models, with GPT-6 Cyber and GPT-Rosalind as specialised models under trusted access.
• Context is 1.05M tokens on current GPT models. GPT-6 Astra, 6.1 Sol, 6.1 Terra, 6 Luna, 6 Sol, 5.6 Sol and 5.5 list a 1,050,000-token context window and 128K max output; gpt-oss-120b lists 131,072.
• Prices run from $0.10 to $20 per million input tokens. GPT-6 Luna is $0.10 input and $0.50 output; GPT-6 Cyber is $20 and $100; Batch and Flex requests cost 50% of standard rates.
• Cyber and life sciences models are restricted. GPT-6 Cyber and GPT-5.6 Cyber are under OpenAI Daybreak trusted access and GPT-Rosalind is for approved organisations.
• Shutdown dates are published. OpenAI's deprecations page lists a notice date, shutdown date and replacement for each model, and gpt-5, o3 and o3-pro shut down on 11 December 2026.

What is OpenAI?

OpenAI develops the GPT family of large language models, the earlier o-series reasoning models and the open-weight gpt-oss models. Its models page groups current models as flagship, specialised and "More models", and they are offered through the OpenAI API, ChatGPT and partner clouds. The current tiers are named Astra, Sol, Terra and Luna, and specialised Cyber and Rosalind models sit under trusted access programmes. Current models use undated API identifiers such as gpt-6.1-sol, while older models carry dated snapshots.

OpenAI publishes a deprecations page with a notice date, shutdown date and replacement for each retiring model, and deprecated models stay callable until their shutdown date. The gpt-oss-120b and gpt-oss-20b models are the only OpenAI models on the pages checked with published weights, under Apache 2.0. Facts are as of 15 November 2026.

Current OpenAI models at a glance

As of 2026-11-15. Source: OpenAI models pages, per-model pages and changelog. Models are listed by release date, newest first. Category is the grouping used on OpenAI's models page.

OpenAI's models page describes GPT-6 Astra as its most capable model for demanding work, GPT-6.1 Sol as near-Astra performance at lower cost, GPT-6.1 Terra as a fast, efficient model for visual reasoning, computer use and writing, and GPT-6 Luna as its most efficient model for high-volume tasks. GPT-6 Astra offers reasoning effort levels low, medium, high, xhigh and max. Other models still listed under "More models" are GPT-5.6 Terra, GPT-5.6 Luna, GPT-5.5 Pro, GPT-5.4, GPT-5.4 Pro, GPT-5.4 mini, GPT-5.2 and GPT-5.2 Pro; chat-latest is listed as a ChatGPT model not recommended for API use. gpt-oss-120b and gpt-oss-20b are published with open weights under Apache 2.0, and OpenAI states 117B total and 5.1B active parameters for gpt-oss-120b, which its model page says fits on a single H100 GPU. See Open-Weight vs Closed Models Explained.

Key facts for the main OpenAI models

As of 2026-11-15. Identical fields for every model; "Not stated by OpenAI" means not stated by OpenAI on the pages checked (2026-11-15).

Positioning text is taken from OpenAI's models page. The flagship and "More models" GPT entries share the same 1.05M context and 128K max output, so the main differences in the table are release date, knowledge cutoff and price, which is covered in the pricing section.

Features stated on the model pages checked

• Reasoning effort: GPT-6 Astra offers low, medium, high, xhigh and max.
• Pro mode: gpt-6.1-sol with reasoning.mode set to pro in the Responses API replaces the earlier Pro models.
• Batch and Flex: 50% of standard rates.
• Fast mode and Ultrafast: Fast mode at 2x standard rates replaced Priority processing on 30 July 2026; an Ultrafast tier is priced for GPT-6 Astra and GPT-6.1 Sol.
• Prompt caching: 30-minute and 12-hour cache writes are priced per model in the pricing table.

OpenAI model families and naming

• Astra: the tier OpenAI describes as most capable (GPT-6 Astra). A legal variant, gpt-6-astra-law, is priced separately.
• Sol: high-capability models at lower cost than Astra (GPT-5.6 Sol, GPT-6 Sol, GPT-6.1 Sol).
• Terra: fast, efficient models (GPT-5.6 Terra, GPT-6.1 Terra).
• Luna: the most efficient tier for high-volume tasks (GPT-5.6 Luna, GPT-6 Luna).
• Specialised: Cyber models for vulnerability research under Daybreak, and Rosalind models for life sciences.
• API IDs: current models use undated IDs such as gpt-6.1-sol; older models carry dated snapshots, for example gpt-5.5-2026-04-23 and o3-2025-04-16. Aliases such as gpt-5.6 route to a named model, gpt-5.6-sol in that case.
• Pro mode: earlier Pro models (gpt-5-pro, o3-pro) are replaced by gpt-6.1-sol with reasoning.mode set to pro in the Responses API, per the deprecations page.
• Deprecated but still available: the all-models page lists gpt-5.4-nano, gpt-5.3-codex, gpt-5.1, gpt-5, gpt-5-mini, gpt-5-nano, gpt-5-pro, o3-pro, o3, gpt-4.1, gpt-4.1-mini, gpt-4o-mini and gpt-4o as deprecated but still callable until their shutdown dates. See LLM Naming and Versioning Explained.

When was each OpenAI model released?

As of 2026-11-15. Dates are from the OpenAI API changelog, in date order.

The latest release is GPT-6 Cyber, on 5 November 2026, under restricted access. The latest generally available release is GPT-6.1 Terra, on 26 October 2026. OpenAI does not state a release date for GPT-Rosalind or gpt-oss-120b on the pages checked.

How much do OpenAI models cost?

As of 2026-11-15. USD per million tokens (MTok), Standard processing. Batch shows input / output. Cache write shows the 30-minute and 12-hour write prices. Source: OpenAI pricing page.

• Long context: on GPT-6 Astra, GPT-6 Sol, GPT-5.6 Sol and GPT-5.5, prompts over 272K input tokens are charged 2x input and cache rates and 1.5x output for the full request.
• Fast mode: 2x the applicable rates, for example GPT-6 Astra at $20 / $100. An Ultrafast tier lists GPT-6 Astra at $60 / $300 and GPT-6.1 Sol at $12 / $60. Fast mode replaced Priority processing on 30 July 2026.
• Data residency: regional processing endpoints carry a 10% uplift for models released on or after 5 March 2026; FedRAMP endpoints also carry a 10% uplift.
• Promotions: GPT-6 Astra for Law promotional pricing runs at least through 31 December 2026, and GPT-5.6 Sol promotional pricing runs until 30 November 2026.

See Model Pricing Explained and How to set spend limits on LLM API keys.

Where are OpenAI models available?

• First-party: the OpenAI API and ChatGPT.
• Cloud platforms: since 1 June 2026, OpenAI models including gpt-5.4 and gpt-5.5 are on Amazon Bedrock through an OpenAI-compatible Responses API endpoint, with availability varying by region. GPT-6 Astra is listed on Bedrock as openai.gpt-6-astra and on Google Cloud as gpt-6-astra in private preview.
• Regional processing: data residency and FedRAMP endpoints carry a 10% uplift, as above.
• Open weights: gpt-oss-120b and gpt-oss-20b are downloadable for self-hosting.

Restricted Cyber and Rosalind models are available only to organisations approved under the programmes in the next section. Running gpt-oss weights yourself moves security responsibility to you; see the Open-Weight Model Security Guide.

Restricted-access OpenAI models

As of 2026-11-15. Source: OpenAI models page, changelog and data controls page.

OpenAI's data controls page states that OpenAI can make specific models ineligible for zero data retention for approved customers, with advance written notice, where needed to address severe risk.

What is the context window of OpenAI models?

GPT-6 Astra, GPT-6.1 Sol, GPT-6.1 Terra, GPT-6 Luna, GPT-6 Sol, GPT-5.6 Sol and GPT-5.5 list a 1,050,000-token context window and 128,000-token max output. gpt-oss-120b lists 131,072 tokens for both. On GPT-6 Astra, GPT-6 Sol, GPT-5.6 Sol and GPT-5.5, prompts over 272K input tokens are billed at long-context rates for the full request. OpenAI does not state the context window of the Cyber and Rosalind models on the pages checked. See Context Windows Explained.

Is any OpenAI model open source?

gpt-oss-120b and gpt-oss-20b are open-weight models under the Apache 2.0 licence. All GPT-6, GPT-5 and o-series models are proprietary and accessed through OpenAI or partner cloud APIs and ChatGPT.

Model cards and system cards

As of 2026-11-15.

The GPT-6 Astra model page lists context window, max output, knowledge cutoff, modalities, reasoning effort levels, pricing rules and cloud IDs. See Model Cards and System Cards Explained.

Benchmarks

As of 2026-11-15. Provider-reported and independent results are kept in separate columns.

OpenAI's model pages checked for this hub give no benchmark scores. See How to Read LLM Benchmarks.

Deprecations and retirements

As of 2026-11-15. Source: OpenAI deprecations page. Fine-tuned versions of a base model follow the base model's shutdown schedule.

The current flagship models list no deprecation date. See the AI Model Retirement Guide.

Data handling and security

• Training use: OpenAI's data controls page states that API data has not been used to train or improve OpenAI models since 1 March 2023 unless the customer explicitly opts in.
• Default retention: abuse monitoring logs are retained for up to 30 days by default; Responses API data is stored for at least 30 days when store is true or omitted.
• Zero data retention (ZDR): subject to prior approval by OpenAI. Batches, Files, Assistants and Conversations are among the endpoints listed as not ZDR-eligible.
• Restricted models: OpenAI can make specific models ineligible for zero data retention for approved customers, with advance written notice, where needed to address severe risk.
• Key handling: generic controls apply to every provider's API: keep keys out of code, scope and rotate them, and set spend limits.

See AI Data Retention and Deletion and the ChatGPT Enterprise Security Guide.

How to choose an OpenAI model: factual criteria

• Access: Cyber and Rosalind models need trusted access approval; flagship models are generally available.
• Retention requirements: if the organisation runs under ZDR, confirm the endpoint and model remain eligible.
• Price per token: compare input, cached, output and batch rates, including the 272K long-context multiplier.
• Context: 1.05M on current GPT models, 131,072 on gpt-oss-120b.
• Lifecycle: check the shutdown date and replacement before building on an older ID.
• Platform: confirm the model is listed in the chosen cloud's catalogue and region.
• Batch and fast tiers: Batch and Flex are 50% of standard rates; Fast mode is 2x and Ultrafast is priced separately.
• Data residency: regional processing and FedRAMP endpoints add 10% for models released on or after 5 March 2026.
• Self-hosting: gpt-oss-120b and gpt-oss-20b are the open-weight options.

Practitioner checklist

• Pin production code to an explicit model ID or snapshot and record it in the AI inventory.
• Review the deprecations page on a schedule to catch new notices.
• Move workloads off gpt-5, o3 and o3-pro before 11 December 2026.
• Check ZDR eligibility for each endpoint before relying on it.
• Re-estimate cost for prompts over 272K tokens.
• Confirm partner cloud availability by region separately from OpenAI's.
• Store OpenAI API keys in a secrets manager and set spend limits. See the LLMjacking Guide.

Standards and references

• OpenAI: Models (accessed 2026-11-15)
• OpenAI: All models (accessed 2026-11-15)
• OpenAI: GPT-6 Astra model page (accessed 2026-11-15)
• OpenAI: Pricing (accessed 2026-11-15)
• OpenAI: Deprecations (accessed 2026-11-15)
• OpenAI: API changelog (accessed 2026-11-15)
• OpenAI: Data controls in the OpenAI platform (accessed 2026-11-15)

Related NHI Mgmt Group resources: OpenAI API Keys Security Guide, Securing OpenAI Agents SDK and Responses API Guide, Reasoning Models Explained, LLM Security Guide.