OpenAIopenai.com

GPT-5.5

GPT-5.5 is OpenAI's newest frontier model, built for the hardest coding and professional work. If your task needs deep reasoning over a lot of context, this is the one to reach for.

Context window: 1,050,000 tokens
Max output: 128,000 tokens
Released: 24 Apr 2026
Knowledge cutoff: Dec 2025
API identifier: gpt-5.5

Pricing · per 1M tokens

Input$5.00

Output$30.00

Cached input$0.50

Cached input is ~10× cheaper — it pays off when you send the same context again and again.

Overview

What GPT-5.5 is good at

OpenAI calls it "a new class of intelligence for coding and professional work." In plain terms, it thinks more carefully than GPT-5.4 and uses fewer tokens to get there. It reads up to 1,050,000 tokens at once, so you can feed it a whole codebase or a stack of filings.

Here's the honest part. It costs $5 per million input tokens and $30 per million output, which is real money at scale. That price buys frontier-level answers, not everyday chat.

So use it where the quality pays off. Think complex financial models, a large legacy codebase migration, or reasoning across dozens of research papers.

For routine drafting, simple support replies, or short classification jobs, a smaller model does the same work for far less. Reach for GPT-5.5 when the problem is genuinely hard. Cached input at $0.50 per million trims the parts you send again and again.

Real numbers

What GPT-5.5 costs in practice

Real enterprise workloads, with the token assumptions shown openly. Your mileage varies — these are honest starting points, not guesses.

FinanceFinance & Accounting

Complex financial modeling assistant

$0.840per request

$336per month

Your analysts hand GPT-5.5 a quarter of filings, prior models, and internal notes, then ask it to draft the assumptions for a new forecast. It reads everything in one pass, and the reasoning holds up across hundreds of numbers. That accuracy is why the higher price earns its place here.

120,000 input tokens8,000 output tokens400 requests / month

Cached input helps a lot here, since the same filings get queried again and again across the month.

ManufacturingInformation Technology

Large legacy codebase migration

$1.35per request

$202.5per month

You point GPT-5.5 at the large legacy codebase that runs your operations and ask it to move to a new framework. It holds many files in context at once, so it tracks how one change ripples through the rest. For a job this tangled, that whole-picture view is worth paying for.

180,000 input tokens15,000 output tokens150 requests / month

A smaller model can handle routine edits, so save GPT-5.5 for the hard cross-file changes.

HealthcareResearch & Innovation

Clinical research synthesis

$0.630per request

$189per month

Your research team gives GPT-5.5 dozens of papers and asks how the findings fit together. It reasons across all of them at once and flags where the evidence disagrees. For high-stakes clinical questions, that careful synthesis justifies the cost.

90,000 input tokens6,000 output tokens300 requests / month

Cached input trims the bill when the same reference papers come up across many questions.

How we work these out: cost = (input ÷ 1M × $5.00) + (output ÷ 1M × $30.00), then × monthly volume. List prices only, no cached-input discount applied — so these are the ceiling, not the floor.

FAQ

Common questions about GPT-5.5

How much does GPT-5.5 cost?+

GPT-5.5 costs $5.00 per million input tokens and $30.00 per million output tokens. If you send the same context again, cached input drops to $0.50 per million. It sits at the premium end, so it's built for hard work rather than everyday chat.

What is GPT-5.5's context window?+

GPT-5.5 has a context window of 1,050,000 tokens. That's big enough to hold a whole codebase or a stack of long documents in a single request. It can write back up to 128,000 tokens in one response.

GPT-5.5 vs GPT-5.4 - which should you use?+

Pick GPT-5.5 for your hardest reasoning and coding work, and keep GPT-5.4 for most everything else. GPT-5.5 is more intelligent and uses fewer tokens to reach an answer, but it's priced higher. So match the model to how hard the task really is.

Is GPT-5.5 good for coding?+

Yes. OpenAI built GPT-5.5 for coding and demanding professional work. Its large context window lets you feed it a whole codebase, so it can reason across many files at once. For big refactors and tricky migrations, that depth earns its higher price.

When is GPT-5.5 overkill?+

GPT-5.5 is overkill for routine work like short replies, simple classification, or everyday drafting. A smaller, lower-priced model handles those just as well for a fraction of the cost. Save GPT-5.5 for problems that genuinely need frontier reasoning or a very large context.

Pricing as of 8 Jul 2026 · reviewed monthlySources: OpenAI API pricing · OpenAI model docs

Not sure GPT-5.5 fits your workload?

We'll size the right model for your use case and give you a real cost estimate. One free call, no slides.

Talk to us