GPT-5.4 Image 2 API: toman pricing and code

What is GPT-5.4 Image 2 good for?

The id for GPT-5.4 Image 2 in our API is "openai/gpt-5.4-image-2", served from OpenAI. Context is 272,000 tokens; past that you have to summarise the history yourself. A single response can run to 128,000 tokens. On price it sits in the "expensive" band — cheaper than 359 and dearer than 47 of the other paid models in the catalogue.

What it can do beyond plain text: it reads images directly, which makes it a real option for invoices, forms and screenshots; it takes files such as PDFs as input; it returns schema-valid JSON through response_format, ready to hand to your code; it has a reasoning mode that pays off on multi-step problems, maths and debugging; it can return generated images. All of it works through the standard parameters of the official OpenAI SDK — no custom client, no wrapper. Keep in mind that reasoning tokens are output tokens and do appear on the bill.

Pricing is 2,321,000 toman per 1M input tokens and 4,351,875 per 1M output tokens. A 1,000-word round trip on GPT-5.4 Image 2 lands near 8,675 toman. It supports cached input: repeated context is billed at 580,250 toman per 1M, which matters a lot if your system prompt is long. Each generated image costs around 8.70 toman. You are always charged for the usage the request actually reported, never for the estimate, and a failed request costs nothing.

The closest alternative with the same capabilities from a different provider is Sonar Pro: GPT-5.4 Image 2 works out roughly 1.3× more expensive, and its context window is larger. Both run on the same key and the same code, so trying the other one is a single string change.

To make the figure concrete: 100,000 toman of credit buys roughly 12 thousand-word requests on GPT-5.4 Image 2, and every 1,000 toman is about 115 words of round trip. A job with one million input tokens and one million output tokens comes to 6,672,875 toman in total. Filling this model's 272,000-token window costs 631,312 toman on the input side alone, which is the real reason to keep conversation history short.

Its nearest relative in the catalogue is "openai/gpt-5-image", and choosing between those two is where most people hesitate. On a thousand-word request this variant works out 1.2× dearer (8,675 against 7,543 toman). The context windows differ too: 272,000 against 400,000 tokens. On parameters, this one takes reasoning_effort while the other takes temperature, top_p.

OpenAI has 95 models in our catalogue; the cheapest is gpt-oss-20b at 60 toman per thousand words and the dearest o1-pro at 282,872. Among the less common parameters it accepts logit_bias, logprobs, reasoning_effort, top_logprobs — all through the standard request body. It does not support temperature, tool_choice, tools, top_p, which most models here do, so test before switching if your code relies on them. By context size the nearest option from another provider is Trinity Large Thinking at 262,144 tokens.

Good fits: work where the quality of the answer matters more than its cost, logic puzzles and code review, pulling text and fields out of images, extracting data against a fixed schema, analysing a long document or codebase in one request.