GPT-5.4

OpenAI's value frontier: the previous flagship, still the right default for most production work — near-GPT-5.5 capability at half the price.

GPT-5.4 is the model many teams should actually be using. It was the flagship before GPT-5.5 arrived, and it remains fully supported as the cost-quality balance point of OpenAI's lineup. At $2.50 in / $15 out per million tokens it's half the price of GPT-5.5, while giving up relatively little for everyday work — strong coding, reasoning, vision, and a large 400K context windowThe maximum amount of text the model can "see" at once — prompt plus prior conversation plus any documents you give it. Measured in tokens (which are roughly three-quarters of a word each). A 128K context window is about 96,000 words of input — roughly a 400-page book. Larger context windows let the model work with bigger documents but cost more to run..

The logic is simple: GPT-5.5 is built for the hardest problems, but most production traffic doesn't need the absolute frontier. For chatbots, document processing, standard coding assistance, and the bulk of business workloads, GPT-5.4 delivers very similar results at a meaningfully lower bill. OpenAI's own guidance is to reserve the flagship for work that truly needs it and run the rest on 5.4 (or cheaper mini/nano tiers).

Like the rest of the GPT line it's closed and API-only, with the same enterprise and Azure options for regulated buyers.

• The default model for most production workloads — strong capability at a reasonable rate.
• Coding assistance, document processing, and chatbots that don't need the absolute frontier.
• Cost-conscious teams that still want OpenAI-tier quality and tooling.
• Large-context tasks within the 400K window.

• The very hardest coding or research problems — step up to GPT-5.5.
• Ultra-high-volume cheap tasks — drop to GPT-5.4 Mini or nano tiers.
• Self-hosting — it's closed; use gpt-oss-120b.
• Buyers chasing the lowest possible tokenThe basic unit of text a model reads and writes. Tokens are roughly three-quarters of a word in English — so 100 tokens is about 75 words. Models don't see letters or words directly; they see tokens. Pricing is almost always quoted per million tokens, and context windows are measured in tokens rather than words. price — open models and DeepSeek go cheaper.

Proprietary and API-only, on the same terms as the rest of the GPT line: commercial rights to your outputs, no rights to the model, accessed through OpenAI's API or Azure under OpenAI's usage policies. Enterprise and zero-retention data options are available. For owning or self-hosting, look at gpt-oss.

Against GPT-5.5, GPT-5.4 gives up some capability and context for half the price — the better choice for most production traffic. Against GPT-5.4 Mini, it's more capable but pricier; route simpler tasks down to mini. Against open alternatives like gpt-oss-120b or Google's Gemma 4 31B, the trade is the familiar closed-vs-open one: managed convenience and tooling versus weightsThe numerical values inside a trained model that encode everything it has learned. A model is, functionally, a giant list of weights — tens of billions of numbers for a mid-sized model, hundreds of billions for a frontier model. "Open-weight" means those numbers are published. "Downloading the weights" means getting the actual file you'd need to run the model yourself. you own and run.