GPT-4

Retired Oct 2026

OpenAI's original GPT-4 model.

Context: 8K tokens
Max output: 8K
Cutoff: Dec 2023
Released: Mar 2023

Visit OpenAI Compare vs other LLMs →

Lyceum

NVIDIA H200 from $2.20/hr, B200 from $3.40/hr, B300 from $3.80/hr

Per-second billing with no minimum commitment or vendor lock-in

Scale to 8,000 GPUs over InfiniBand for distributed AI training

View Pricing

Capabilities

What GPT-4 accepts and can do, as published by OpenAI.

Text Accepts and generates natural-language text.

Common questions

What is GPT-4 good for?

Text-only work that needs the original GPT-4 for reproducibility. OpenAI calls it an older model and recommends newer ones for new work.

When is GPT-4 not a good fit?

Image input, tool calling and structured output, since it has none of them. It sits near the top of OpenAI's price range, and OpenAI plans to shut down the API model in October 2026.

How does GPT-4 compare with GPT-4 Turbo?

OpenAI calls GPT-4 Turbo the next generation of GPT-4, designed as a cheaper, more capable version of that model. GPT-4 came out 13 months earlier, has a smaller context (8K against 128K tokens) and a larger max output (8K against 4K tokens) and takes no image input, which GPT-4 Turbo does.

Can I self-host GPT-4?

No. OpenAI does not publish the weights; GPT-4 is available only through hosted APIs.