All models

GPT-5.4 Pro

Prompt GPT-5.4 Pro

Languagev1.0.04 inputsTEXT

Give a concise argument for testing public behavior.

Test public behavior because it’s the contract users depend on. If tests target internals, they become brittle: harmless refactors break tests even when the software still works. By testing observable inputs/outputs and side effects, you verify real correctness while preserving freedom to change the implementation.

Try it right here

Fill the inputs the node publishes and run it on your own account. The same ports, defaults and types the studio uses.

Inputs

TEXT
IMAGE
TEXT
INT
Sign in to run

Running a model executes it on your account. Browsing the catalog stays free and needs no account.

Output

Fill in the inputs and run the node to see its output here.

Call it from your code

The same node, called by id from any surface. Every tab pins this node and passes the ports below.

import { BlitClient } from "@blitflow/sdk";

const { runNode } = new BlitClient({ apiKey: process.env.BLITFLOW_TOKEN });

const { outputs } = await runNode("gpt-5-4-pro", {
  prompt: { value: "example" },
  image: { ref: "https://..." },
  systemPrompt: { value: "example" },
  maxTokens: { value: 1024 },
});

Schema

Every port this node publishes, with its type, its hint and the default it runs with. gpt-5-4-pro is the id you call.

Inputs

prompt*TEXT
imageIMAGE
systemPromptTEXT
maxTokensINT1024

Outputs

textTEXT

Chain it with the rest of the catalog

Open the studio