GPT‑5 is an OpenAI reasoning model released on August 7, 2025 for coding, agentic work, and complex problem-solving. It was a flagship model at launch, but OpenAI now describes it as a previous model and recommends GPT‑5.6 for new workloads.

What is GPT‑5?

GPT‑5 is available through the OpenAI API under the model ID gpt-5. It is not the same thing as ChatGPT: ChatGPT is a product that can use different models, while GPT‑5 is a specific model with its own limits, features, and API pricing.

What can GPT‑5 do?

GPT‑5 supports configurable reasoning effort at minimal, low, medium, and high levels. It also supports coding, function calling, structured outputs, streaming, and image input alongside text. Its output is text. Audio and video are not directly supported according to the official model page.

Context window and output

The model has a 400,000-token context window and a maximum output of 128,000 tokens. This makes it useful for long documents and codebases, but a large context window does not guarantee that every detail will be recalled accurately.

GPT‑5 API pricing

OpenAI lists standard pricing of $1.25 per million input tokens, $0.125 per million cached-input tokens, and $10 per million output tokens. These are API prices, not ChatGPT subscription prices, and they can change.

Strengths

  • Documented strength in coding and agentic tasks.
  • A 400,000-token context window.
  • Image input, tool use, and structured outputs.

Limitations

  • OpenAI no longer recommends it as the starting model for new work.
  • It does not directly support audio or video.
  • Important outputs still require human verification.

GPT‑5 and ChatGPT: what is the difference?

ChatGPT is the user-facing application, while GPT‑5 is a model that developers can call through the API. You should not assume that every ChatGPT conversation runs on GPT‑5 or that API specifications apply to every ChatGPT experience.

In short

GPT‑5 remains an important model in OpenAI’s history and may still support existing applications. For new projects, OpenAI recommends beginning with its latest GPT‑5.6 models.