OpenAI models: Every model (including GPT-6) and what it's best for

Keeping track of all the new AI models getting released at the moment is practically a full-time job. The most recent model, GPT-6 Astra, was released about two months after the GPT-5.6 series, which itself was released less than three months after GPT 5.5—you get the picture.

I've been writing about OpenAI's models for the past few years, and it feels like every time I publish an article, another new model drops. OpenAI seems to be serious about pushing point-releases more frequently rather than waiting longer between frontier models and is quite happy for different models to have different version numbers.

So, let's try to simplify things: consider this your complete guide to all of OpenAI's major models. We'll look at what each one is good for, how much they cost, and how to make the most of them in your work. 

But first, I have to apologize for the crimes against consistent naming conventions, capitalization, and hyphenation we're all going to have to endure. GPT-6 Astra, GPT-5.4 nano, gpt-oss-120b, and GPT Image 2.5 are a few of OpenAI's models. Some of them even do much the same thing. Hang in there.

Table of contents:

• Why are there so many OpenAI models?

• Major OpenAI models

• Legacy OpenAI models

• How to choose the right OpenAI model

• Automate OpenAI models

Why are there so many OpenAI models?

OpenAI is one of the most important AI companies. Its ChatGPT chatbot and DALL·E 2 image generator essentially kickstarted the current AI boom. Over the past few years, it's continued to develop large language models, multimodal models, and text-to-image models.

OpenAI is moving fast. As the technology advances, new models have come out, and we've naturally been left with an extensive list—including some legacy models.

OpenAI models at a glance

GPT-6 Astra

GPT-6 Astra is the most powerful model OpenAI has publicly released. It's OpenAI's equivalent to Claude Fable and is truly state-of-the-art. It excels at computer use, software engineering, cybersecurity, science, and professional knowledge work. It's capable of long agentic workflows: give it a task, and let it keep working until it's done. 

The catch with GPT-6 Astra is the price. It costs 2.5x more than GPT-5.6 Sol for a marginal bump in performance in most use cases. It's an incredibly powerful model, but it's far from the default option in most situations—unless you have a very generous company token policy.

With every new model, Zapier tests on hundreds of real workflows, and GPT-6 Astra was a top performer on messy, multi-step workflows where accuracy matters most.

• GPT-6 Astra inputs: Text, images

• GPT-6 Astra outputs: Text

• GPT-6 Astra context window: 1 million tokens

• GPT-6 Astra $10/M tokens(< 272k context); $20/M tokens (> 272 context)

• GPT-6 Astra $50/M tokens (< 272k context); $75/M tokens (> 272 context)

GPT-5.6 Sol

Best OpenAI model for advanced reasoning and logic

GPT-5.6 Sol is the most powerful model in the GPT-5.X series. The series was released less than three months after GPT-5.5, but it was still a significant step up in benchmark performance.

Reasoning models like GPT-5.6 are best for logical, technical, and scientific tasks. If you need the most powerful model OpenAI has for building agentic tools, generating computer code, analyzing text documents and images, or solving complex puzzles, but don't want to pay Astra prices, GPT-5.6 Sol is the one to use.

It costs twice as much to use GPT-5.6 Sol compared to the Terra version of the model and five times more compared to the Luna version. Unless you know you need its cutting edge performance, it's not the best model to start with.

• GPT-5.6 Sol inputs: Text, images

• GPT-5.6 Sol outputs: Text

• GPT-5.6 Sol context window: 1 million tokens

• GPT-5.6 Sol input pricing: $4/M tokens (< 272k context); $8/M tokens (> 272 context)

• GPT-5.6 Sol output pricing: $20/M tokens (< 272k context); $30/M tokens (> 272 context)

GPT-5.6 Terra

Best OpenAI model for balancing reasoning and cost

GPT-5.6 Terra is the middle-ground between the most powerful Sol model and the lighter weight and more affordable Luna model. OpenAI says it's equivalent to the mini version of previous model releases.

• GPT-5.6 Terra inputs: Text, images

• GPT-5.6 Terra outputs: Text

• GPT-5.6 Terra context window: 1 million tokens

• GPT-5.6 Terra input pricing: $2/M tokens(< 272k context); $4/M tokens (> 272 context)

• GPT-5.6 Terra output pricing: $12/M tokens (< 272k context); $18/M tokens (> 272 context)

GPT-5.6 Luna

Best OpenAI model for affordable reasoning and logic

GPT-5.6 Luna is a faster and more affordable alternative to GPT-5.6 Sol and Terra. For well-defined or chatbot responses, it offers a great balance between power and performance. OpenAI says it's equivalent to the nano tier of previous model releases.

• GPT-5.6 Luna inputs: Text, images

• GPT-5.6 Luna outputs: Text

• GPT-5.6 Luna context window: 1 million tokens

• GPT-5.6 Luna input pricing: $0.20/M tokens(< 272k context); $0.40/M tokens (> 272 context)

• GPT-5.6 Luna output pricing: $1.20/M tokens (< 272k context); $1.80/M tokens (> 272 context)

GPT-5.6 Cyber

Best OpenAI model for cybersecurity

GPT-5.6 Cyber is a version of GPT-5.6 Sol specially trained for cybersecurity tasks. It's only available to approved researchers and organizations through the OpenAI Daybreak program. It's also OpenAI's most expensive model.

• GPT-5.6 Cyber inputs: Text, images

• GPT-5.6 Cyber outputs: Text

• GPT-5.6 Cyber context window: 400,000 tokens

• GPT-5.6 Cyber input pricing: $12.50/M tokens

• GPT-5.6 Cyber output pricing: $75/M tokens

GPT-5, GPT-5.1, GPT-5.2, GPT 5.4, and GPT-5.5

Superseded

GPT 5 launched in August 2025, followed by GPT 5.1 in November. GPT-5.2 was a huge step up for OpenAI when it was released in December 2025, but with GPT-5.4 coming on the scene just over two months later (days after GPT-5.3), it was quickly superseded. A few months later, we had GPT-5.5, and they've since all been superseded by GPT 5.6. Many of these models are still available in the API (there are Pro versions of them available too). GPT-5.4 pricing is below for context, but check out the API docs to see the accurate pricing of all the older models.

• GPT-5.4 inputs: Text, images

• GPT-5.4 outputs: Text

• GPT-5.4 context window: 1M token

• GPT-5.4 input pricing: $2.50/M tokens

• GPT-5.4 output pricing: $15/M tokens

GPT-Live-1

Best ChatGPT model for audio

OpenAI recently updated the audio model in ChatGPT to GPT-Live-1 and GPT-Live-1-mini. It's OpenAI's first audio model that can both listen and speak at the same time, so it allows for real-time conversations. It uses GPT-5.5 in the background and will come to the API in the near future. 

• GPT-Live-1 inputs: Audio

• GPT-Live-1 outputs: Text, audio

• GPT-Live-1 pricing: ChatGPT only

GPT-Audio-1.5

Best OpenAI API model for audio

OpenAI's first generally available audio model, it supersedes GPT-Audio and the old GPT-4o Audio model. It can transcribe audio into text or reply with speech. 

• GPT-Audio-1.5 inputs: Text, audio

• GPT-Audio-1.5 outputs: Text, audio

• GPT-Audio-1.5 context window: 128,000 tokens

• GPT-Audio-1.5 input pricing: $2.50/M tokens for text; $32/M tokens for audio

• GPT-Audio-1.5 output pricing: $10/M tokens for text; $64/M tokens for audio

GPT-Realtime-2.1

Best OpenAI model for realtime audio

GPT-Realtime-2.1 is a realtime audio model that allows for naturalistic conversations. It supports reasoning and can use tools for advanced audio workflows. 

• GPT-Realtime-2.1 inputs: Text, audio, image

• GPT-Realtime-2.1 outputs: Text, audio

• GPT-Realtime-2.1 context window: 128,000 tokens

• GPT-Realtime-2.1 input pricing: $4/M tokens for text; $32/M tokens for audio

• GPT-Realtime-2.1 output pricing: $24/M tokens for text; $64/M tokens for audio

GPT-Realtime-2.1 Mini

Best OpenAI model for affordable realtime audio

GPT-Realtime-2.1 Mini is a realtime audio model that allows for naturalistic conversations. It's more affordable than GPT-Realtime-2.1 but still supports reasoning tool use.

• GPT-Realtime-2.1 Mini inputs: Text, audio, image

• GPT-Realtime-2.1 Mini outputs: Text, audio

• GPT-Realtime-2.1 Mini context window: 128,000 tokens

• GPT-Realtime-2.1 Mini input pricing: $0.60/M tokens for text; $10/M tokens for audio

• GPT-Realtime-2.1 Mini output pricing: $2.40/M tokens for text; $20/M tokens for audio

gpt-oss-120b

Best open weight OpenAI model

gpt-oss-120b is OpenAI's most powerful open weight model, and it's the most powerful open model that can run on a single H100 GPU. It's released under an Apache 2.0 license, so you can download it and run it on your own server or choose from any third-party inference provider. You can also fine-tune the model to suit your needs. This is big for OpenAI, which historically, has only released proprietary models.

gpt-oss-120b has 117 billion parameters. It uses a mixture-of-experts architecture, so only 5.1 billion parameters are active at once. 

• gpt-oss-120b inputs: Text

• gpt-oss-120b outputs: Text

• gpt-oss-120b context window: 131,072 tokens

• gpt-oss-120b input pricing: Varies by provider

• gpt-oss-120b output pricing: Varies by provider

gpt-oss-20b

Best medium open weight OpenAI model

gpt-oss-20b is OpenAI's medium open weight model. It's designed to run locally and can even run on a sufficiently powerful laptop. Like gpt-oss-120b, it's released under an Apache 2.0 license, so you can download it yourself or choose from any third-party inference provider, and you can fine-tune the model to suit your needs. 

gpt-oss-20b has 21 billion parameters. It uses a mixture-of-experts architecture so only 3.6 billion parameters are active at once. 

• gpt-oss-120b inputs: Text

• gpt-oss-120b outputs: Text

• gpt-oss-120b context window: 131,072 tokens

• gpt-oss-120b input pricing: Varies by provider

• gpt-oss-120b output pricing: Varies by provider

GPT Image 2.5

Best OpenAI model for generating images

GPT Image 2.5 is the successor to OpenAI's DALL·E image models, and it's one of the best text-to-image models available. It supersedes GPT Image 1, GPT Image 1.5, and GPT Image 2.

OpenAI's pricing for image models is quite complicated because images are converted into tokens based on their size and quality.

• GPT Image 2.5 inputs: Text, images

• GPT Image 2.5 outputs: Images

• GPT Image 2.5 input pricing: $5/M text tokens; $8/M image tokens

• GPT Image 2.5 output pricing: $30/M image tokens

Legacy OpenAI models

OpenAI has developed and deprecated dozens of models over the past few years. Here are some of the major models that are no longer supported, actively used, or state-of-the-art:

• o3

• o3-pro

• o4-mini

• o1

• o1-pro

• GPT-5.5 Pro

• GPT-5.5

• GPT-5.4 Pro

• GPT-5.4

• GPT-5.4 mini

• GPT-5.4 nano

• GPT-5.3 Codex

• GPT-5.2

• GPT-5.1

• GPT-5

• GPT-5 mini

• GPT-5 nano

• GPT-4.5

• GPT-4.1

• GPT-4.1 mini

• GPT-4.1 nano

• GPT-4o

• GPT-4o mini

• GPT-4 Turbo

• GPT-4

• GPT-3.5 Turbo

• GPT-3

• gpt-audio

• gpt-audio-mini

• GPT-4o Audio

• gpt-realtime

• gpt-realtime-mini

• Whisper

• GPT Image 2

• GPT Image 1.5

• GPT Image 1 mini

• GPT Image 1

• DALL·E 3

• DALL·E 2

• Sora 2 Pro

• Sora 2

• Sora 1

While most of these have been replaced or superseded by newer models, it's noteworthy that there's been no announced replacement for OpenAI's now-deprecated Sora text-to-video models. This may change in the coming months.

How to choose the right OpenAI model

OpenAI has lots of models, but they each serve a purpose. GPT-6 Astra and the various GPT-5.6 models offer a different balance between price, performance, and speed. There's certainly a lot of overlap between the different models' abilities, especially when you take into account older models that haven't yet been replaced. But in general, one should strike you as the best balance between power, price, and ongoing development.

OpenAI keeps releasing new models, changing pricing, and otherwise mixing things up. I'll do my best to keep this list as current and accurate as possible in the meantime.

Use OpenAI models in your existing workflows

OpenAI's models are most useful when they're woven into the work you're already doing—not sitting in a separate chat window you have to context-switch to. Zapier connects GPT-6 and the rest of OpenAI's model lineup to 9,000 apps, so you can add state-of-the-art reasoning to all your workflows.

With AI by Zapier, you can add AI steps to your deterministic workflows—some models are built in, so you don't even need an OpenAI account. That means things like automatically triaging support tickets, generating CRM summaries when a deal moves stages in Salesforce, or drafting a Slack message whenever a form submission comes in. And because Zapier handles OAuth for every connected app, your credentials never touch the model's context.

Or you can install Zapier in any AI assistant—like Claude, ChatGPT, or Cursor—or the terminal with a single command. From there, your AI can take real actions across your stack without ever leaving the chat window. Learn more about how to use OpenAI models with Zapier.

Related reading:

• How to create a custom GPT: A beginner's guide

• What is mixture of experts in AI?

• What is AI orchestration? A guide to intelligent systems

• What is gpt-oss?

• What is deterministic AI?

• Prevent lock-in with AI model flexibility on Zapier

This article was originally published in May 2025. The most recent update was in September 2026.