This website uses cookies

Read our Privacy policy and Terms of use for more information.

Last Updated: November 17, 2026

Summary: OpenAI's current API flagship models are GPT-6 Astra ($10 input / $50 output per 1M tokens), GPT-6.1 Sol ($2 / $10), GPT-6.1 Terra ($1 / $5) and GPT-6 Luna ($0.10 / $0.50), and all four share a 1.05M-token context window and 128K max output. Alongside them sit GPT-Image-2.6 for images, GPT-Realtime-2.1 and GPT-Live-1 for voice, GPT-Transcribe for speech to text, text-embedding-3 for search, a free moderation model and the open-weight gpt-oss models. In ChatGPT, the GPT-6 option runs on GPT-6 Sol, with GPT-6 Pro (built on Astra) for Pro, Business and Enterprise users.

OpenAI has replaced most of its lineup since September 2026. GPT-5, o3, GPT-4o, GPT-4.1 and the original Whisper and TTS models are all either shut down or on a published retirement schedule, and many lists online still recommend them.

This OpenAI models list covers every current model with official prices and limits, the ChatGPT models by plan and a full retirement calendar, all checked against OpenAI's documentation in November 2026.

Free download: The AI Stats Sheet 2026. 36 key AI numbers for business on four printable pages, with the source under every figure. Join 2,000+ business leaders who read AI Business Weekly and get it in your inbox right away.

In this guide

What Are the Current OpenAI API Models?

OpenAI's current API models are led by four flagship GPT-6 models: Astra, 6.1 Sol, 6.1 Terra and Luna. All four take text and image input, return text, and support reasoning, according to OpenAI's models page, and the individual model pages list the Responses API, Chat Completions and Batch as supported endpoints.

Rows of servers in a data center, the kind of infrastructure that runs OpenAI API models

Every GPT-6 flagship runs on the same 1.05M-token context window. Photo: Planet Volumes / Unsplash+

Model (API ID)

Context window

Max output

Input / cached / output per 1M tokens

Knowledge cutoff

Best for

GPT-6 Astra (gpt-6-astra)

1.05M

128K

$10 / $1 / $50

Apr 30, 2026

The hardest reasoning, coding, research and computer-use work

GPT-6.1 Sol (gpt-6.1-sol)

1.05M

128K

$2 / $0.10 / $10

Apr 30, 2026

Near-Astra quality on complex work at a fifth of the price

GPT-6.1 Terra (gpt-6.1-terra)

1.05M

128K

$1 / $0.05 / $5

Jun 1, 2026

Fast visual reasoning, precise computer use and natural writing

GPT-6 Luna (gpt-6-luna)

1.05M

128K

$0.10 / $0.01 / $0.50

May 18, 2026

High-volume tasks such as classification, extraction and routing

Prices are OpenAI's Standard tier rates from its official pricing page. The Astra, 6.1 Sol and Luna model pages cap input at 922,000 tokens inside the 1.05M window. All four support Functions, Web search, File search and Computer use as tools.

OpenAI's API changelog shows GPT-6 Astra launched September 3, 2026, GPT-6 Sol and GPT-6 Luna on September 22, GPT-6.1 Sol on September 29 and GPT-6.1 Terra on October 26. For the generation these replaced, see what GPT-5 is.

How much does each flagship model cost on a real task?

The price gap between OpenAI's flagship models is 100x from top to bottom. Take a job that sends 1 million input tokens and gets 200,000 output tokens back:

  • GPT-6 Astra: $10 input + $10 output = $20

  • GPT-6.1 Sol: $2 + $2 = $4

  • GPT-6.1 Terra: $1 + $1 = $2

  • GPT-6 Luna: $0.10 + $0.10 = $0.20

Prompt caching changes that math for repeated context. If 80% of the input on GPT-6.1 Sol is a cached system prompt or document, the same job costs $0.40 for fresh input, $0.08 for cached input and $2 for output, or about $2.48 instead of $4 (cache-write charges not included). Our full OpenAI API pricing guide covers Batch, Flex, Fast mode and cache-write rates in detail.

Does long context cost more?

Yes. OpenAI's model pages for all four flagships state that prompts over 272K input tokens are billed at 2x the input and cache rates and 1.5x the output rate for the whole request. A GPT-6.1 Sol call with 400K input tokens and 20K output tokens costs $1.90 at those rates, versus $1.00 at short-prompt rates. Our guide to what a context window is explains the basics.

Which OpenAI Models Handle Images, Voice and Transcription?

OpenAI's current non-text models are GPT-Image-2.6 and 2.5 for images, GPT-Realtime-2.1, GPT-Live-1 and GPT-Audio-1.5 for voice, and GPT-Transcribe for speech to text. Each uses its own endpoints.

Model (API ID)

What it does

Official price

Limits

GPT-Image-2.6 Sunburst (gpt-image-2.6-sunburst)

Detailed image generation and editing

Text in $5, image in $8, image out $30 per 1M tokens

Image generation and edit endpoints only

GPT-Image-2.6 Flare (gpt-image-2.6-flare)

Fast image iteration

Same token rates as Sunburst on the pricing page

See model page

GPT-Image-2.5 Sunburst and Flare

Previous image generation, released September 8, 2026

See pricing page

Still listed as current

GPT-Realtime-2.1 (gpt-realtime-2.1)

Speech-to-speech reasoning model with tool use

Audio $32 in / $64 out; text $4 / $24 per 1M tokens

128K context, 32K max output

GPT-Realtime-2.1 mini (gpt-realtime-2.1-mini)

Cheaper, faster voice agents

Audio $10 / $20; text $0.60 / $2.40 per 1M tokens

128K context, 32K max output

GPT-Live-1 (gpt-live-1)

Natural voice conversations with interruption handling

$0.05 per minute, billed per second

Backend model use billed separately

GPT-Transcribe (gpt-transcribe)

High-accuracy transcription of files and realtime turns

$0.0045 per minute

Accepts keyword and language hints

GPT-Live-Transcribe (gpt-live-transcribe)

Low-latency streaming transcription

$0.017 per minute

Realtime use

GPT-Realtime-Translate (gpt-realtime-translate)

Streaming speech-to-speech translation

$0.034 per minute

Realtime use

OpenAI also still lists GPT-Realtime-2, GPT-Realtime-1.5, GPT-Realtime-Whisper and GPT-Audio-1.5 as current. GPT-Image-2.6 arrived November 4, 2026, and the older gpt-image-2 shuts down February 10, 2027.

OpenAI has no video model in its API now. The deprecations page shows the Videos API, sora-2 and sora-2-pro were scheduled to shut down on September 24, 2026, with "no one-to-one replacement."

Does OpenAI Still Offer Embeddings, Moderation and Open-Weight Models?

Yes. OpenAI's current catalog still includes three embedding models, one free moderation model and two open-weight gpt-oss models you can download and run yourself.

Model

Type

Price

Key detail

text-embedding-3-large

Embeddings

$0.13 per 1M tokens

The larger text-embedding-3 model

text-embedding-3-small

Embeddings

$0.02 per 1M tokens

OpenAI calls it an improved version of its ada model

text-embedding-ada-002

Embeddings

See pricing page

Older model, still listed

omni-moderation-latest

Moderation

Free

Flags harmful text and images

gpt-oss-120b

Open-weight reasoning

Free to download (Apache 2.0)

117B parameters (5.1B active), 131,072-token context, built for a single H100 GPU

gpt-oss-20b

Open-weight, low latency

Free to download (Apache 2.0)

21B parameters (3.6B active), 131,072-token context

OpenAI's gpt-oss model pages show zero default rate limits on its paid API tiers, so treat gpt-oss as a model you host yourself.

Which specialized OpenAI models need approval?

OpenAI prices its specialized models separately. GPT-6 Cyber (gpt-6-cyber), released November 5, 2026, is for vulnerability research under OpenAI's Daybreak program, which needs setup before a first request, at $20 input and $100 output per 1M tokens. GPT-5.6 Cyber is also still listed. For life sciences, GPT-6 Rosalind (gpt-6-rosalind) costs $15 and $75 and, per the changelog, is open only to organizations approved through OpenAI's trusted-access program. The models page also lists the earlier GPT-Rosalind (gpt-rosalind-research) for approved organizations. GPT-6 Astra for Law (gpt-6-astra-law) is $10 and $50, with promotional pricing at least through December 31, 2026.

Which Models Are in ChatGPT by Plan?

ChatGPT uses different model names from the API. Free and Go users get GPT-5.6 Luna by default, while Plus, Pro, Business and Enterprise users can pick GPT-6, which runs on GPT-6 Sol, according to OpenAI's help article GPT-6 and other models in ChatGPT.

Two people working on laptops at a shared desk, representing teams choosing ChatGPT plans

Which model you get in ChatGPT depends on your plan. Photo: Andrej Lišakov / Unsplash+

ChatGPT plan

Default or main model

Thinking levels included

GPT-6 Pro (on Astra)

Free and Go

GPT-5.6 Luna, which also powers Think

Not Medium, High or Extra High

No

Plus

GPT-6

Medium and High

No (Astra available in ChatGPT Work and Codex)

Pro ($100 and $200)

GPT-6

Medium, High, Extra High, Pro

Yes, with a weekly limit

Business

GPT-6

Medium, High, Extra High, Pro

Yes, admins can restrict models

Enterprise

GPT-6

Medium, High, Extra High, Pro

Yes, subject to workspace permissions

In ChatGPT, GPT-6 replaced "Latest" in the model menu, and a Thinking slider runs from Instant to Pro. GPT-6 Sol powers Instant through Extra High, and GPT-6 Pro, built on GPT-6 Astra, powers Pro. GPT-5.6 Sol and GPT-5.6 Sol Pro remain selectable, while Terra and Luna are not selectable in standard chat. Enterprise and Edu limits are covered in a separate help article.

Usage caps differ by plan. Free and Go include unlimited everyday text chats, subject to abuse safeguards, with separate limits on uploads, image generation and voice. Pro does not include unlimited GPT-6 Pro use: on the $200 plan, hitting the weekly limit drops you to GPT-5.6 Thinking at Medium. Our ChatGPT pricing guide covers what each plan costs.

What about ChatGPT Work and Codex?

OpenAI's ChatGPT Learn models page recommends Astra, GPT-6.1 Sol, GPT-6.1 Terra and Luna for Work and Codex. GPT-6.1 Sol uses a default context window of 500,000 tokens there, less than half its API window. With GPT-5.5 retired from Codex, OpenAI points paid users to GPT-6.1 Sol and Free and Go users to GPT-6 Luna in the desktop app.

Can you use the ChatGPT model in the API?

Yes, through chat-latest, which OpenAI describes as the "latest Instant model used in ChatGPT." It has a 400,000-token context window and costs $5 input, $0.50 cached and $30 output per 1M tokens. OpenAI's catalog lists it under ChatGPT models "not recommended for API use," and its model page recommends GPT-6 Astra for production.

Which Older OpenAI Models Can You Still Use in the API?

OpenAI's full model catalog lists these "more models" that still work but are no longer the recommended picks:

  • GPT-6 Sol (gpt-6-sol), followed by GPT-6.1 Sol one week later

  • GPT-5.6 Sol, Terra and Luna, released July 9, 2026

  • GPT-5.5 and GPT-5.5 Pro

  • GPT-5.4, GPT-5.4 Pro and GPT-5.4 mini

  • GPT-5.2 and GPT-5.2 Pro

None of these has a published shutdown date on OpenAI's deprecations page as of this update.

Free download: The AI Stats Sheet 2026. 36 key AI numbers for business on four printable pages, with the source under every figure. Join 2,000+ business leaders who read AI Business Weekly and get it in your inbox right away.

Which OpenAI Models Are Being Retired?

OpenAI is retiring GPT-5, GPT-5 mini, GPT-5 nano, GPT-5 Pro, o3 and o3-pro on December 11, 2026, and GPT-4o, GPT-4o mini, GPT-4.1 and GPT-4.1 mini snapshots on January 14, 2027. Every upcoming shutdown:

Shutdown date

Models or features retiring

OpenAI's recommended replacement

Nov 30, 2026

Reusable prompts (/v1/prompts), Evals, Agent Builder

Prompts in app code; Promptfoo; Agents SDK

Dec 1, 2026

gpt-image-1-mini, gpt-image-1.5, chatgpt-image-latest

GPT-Image-2.6 Sunburst or Flare

Dec 11, 2026

gpt-5, o3, gpt-5-pro, o3-pro

GPT-6.1 Sol (pro reasoning mode for the Pro models)

Dec 11, 2026

gpt-5-mini / gpt-5-nano

GPT-6.1 Terra / GPT-6 Luna

Jan 6, 2027

tts-1, tts-1-hd, gpt-4o-mini-tts

GPT-Realtime-2.1 mini

Jan 14, 2027

Dated gpt-4o, gpt-4o-mini, gpt-4.1 and gpt-4.1-mini snapshots

GPT-6 Luna

Jan 20, 2027

gpt-realtime, gpt-realtime-mini, gpt-audio, gpt-audio-mini and GPT-4o realtime/audio models

GPT-Realtime-2.1 (or mini); GPT-Audio-1.5

Feb 1, 2027

gpt-5.5-cyber-preview

Most capable cyber model available to you

Feb 10, 2027

gpt-image-2

GPT-Image-2.6 Sunburst or Flare

Feb 26, 2027

whisper-1, gpt-4o-transcribe, gpt-4o-mini-transcribe, gpt-4o-transcribe-diarize

GPT-Transcribe or GPT-Live-Transcribe

Apr 1, 2027

gpt-5.1, gpt-5.3-codex / gpt-5.4-nano

GPT-6 Sol / GPT-6 Luna

Which OpenAI models already shut down in 2026?

OpenAI's deprecations page set an October 23, 2026 shutdown for GPT-3.5 Turbo, GPT-4, GPT-4 Turbo, o1, o1-pro, o3-mini, o4-mini, GPT-4.1 nano and gpt-image-1. DALL-E 2 and DALL-E 3 were set to shut down May 12, 2026, chatgpt-4o-latest on February 17, 2026, and the deep research models o3-deep-research and o4-mini-deep-research on July 23, 2026. The Assistants API shut down on August 26, 2026, per the changelog.

Swapping the model name is usually the main fix, but OpenAI's latest-model guide flags real differences. GPT-6 Astra and GPT-6.1 Sol do not accept the none reasoning effort, and requests with reasoning on should drop temperature, top_p and logprobs settings.

How Do You Check Which OpenAI Models Your API Key Can Use?

Call the List Models endpoint, GET /v1/models, which OpenAI's API reference says "lists the currently available models" with basic details such as the owner. Access varies by usage tier and program. OpenAI simplified its usage tiers from five to three (Build, Launch and Grow) on October 6, 2026, and the GPT-6 Luna model page lists the Free tier as "not supported."

Developer coding on multiple monitors in a dark room, checking which models an API key can access

Your usage tier decides which models your API key can call. Photo: Paris Bilal / Unsplash+

Which OpenAI Model Should You Choose?

For most business workloads, start with GPT-6.1 Sol, which OpenAI positions as near-Astra quality at a fifth of the price, and move up or down only when testing shows a reason.

  • Pick GPT-6 Astra for complex agents, hard coding problems and long research tasks.

  • Pick GPT-6.1 Sol as the everyday default for chat assistants, analysis and coding.

  • Pick GPT-6.1 Terra for screenshot and chart reading, computer-use agents and writing where speed matters.

  • Pick GPT-6 Luna for anything you run millions of times: tagging, extraction, routing and simple replies.

Before you commit, compare our Claude API pricing and Gemini API pricing guides, or see our OpenAI statistics.

Frequently Asked Questions

What is the latest OpenAI model?
GPT-6 Astra is OpenAI's most capable model, released September 3, 2026, and its newest flagship release is GPT-6.1 Terra, released October 26, 2026. The newest models overall are GPT-6 Cyber (November 5) and GPT-Image-2.6 Sunburst and Flare (November 4).

What models are available in the OpenAI API?
The current flagship OpenAI API models are GPT-6 Astra, GPT-6.1 Sol, GPT-6.1 Terra and GPT-6 Luna. OpenAI also offers GPT-Image-2.6 and 2.5 for images, GPT-Realtime-2.1 and GPT-Live-1 for voice, GPT-Transcribe for speech to text, text-embedding-3 models, omni-moderation and the open-weight gpt-oss models.

Which ChatGPT models are free?
ChatGPT Free and Go users get GPT-5.6 Luna as the default model, which also powers the Think option. GPT-6 is for paid plans, and GPT-6 Pro is limited to Pro, Business and Enterprise.

Is GPT-4o still available in the OpenAI API?
Not for long. OpenAI's deprecations page sets January 14, 2027 as the shutdown date for the remaining dated GPT-4o, GPT-4o mini, GPT-4.1 and GPT-4.1 mini snapshots, and recommends GPT-6 Luna as the replacement.

What is the context window of OpenAI's models?
All four GPT-6 flagship models have a 1.05M-token context window and 128,000 max output tokens, with input capped at 922,000 tokens on Astra, 6.1 Sol and Luna. GPT-Realtime-2.1 has 128K, chat-latest has 400K, and the gpt-oss models have 131,072 tokens.

The Bottom Line

OpenAI's API lineup now comes down to four GPT-6 text models priced from $0.10 to $10 per million input tokens, plus dedicated image, voice, transcription, embedding and moderation models. GPT-6.1 Sol is the sensible default, Luna the cost play and Astra the quality pick.

The bigger job is migration: GPT-5 and o3 shut down December 11, 2026, and GPT-4o, GPT-4.1, TTS and Whisper follow by February 2027.

Related Articles

⭐ Get more AI Business Weekly in Google: add us as a preferred source on Google.

Written by Sameer Khan with AI assistance, fact-checked against the sources linked above.