What is GPT-5.4 designed to do best?

OpenAI positions GPT-5.4 as a model for complex professional work, especially documents, spreadsheets, presentations, coding, and tool-using workflows. In ChatGPT, OpenAI says GPT-5.4 Thinking is stronger at spreadsheet editing, polished frontend code, slideshow creation, and document understanding.

Is GPT-5.4 available in Canada?

Yes. OpenAI launched GPT-5.4 across ChatGPT, the API, and Codex on 5 March 2026, and the company separately said its ChatGPT for Excel beta is rolling out in Canada for Business, Enterprise, Edu, Teachers, Pro, and Plus users.

How much does GPT-5.4 cost through the API?

OpenAI lists gpt-5.4 at $2.50 per million input tokens, $0.25 per million cached input tokens, and $15 per million output tokens. gpt-5.4-pro is listed at $30 per million input tokens and $180 per million output tokens.

Should teams replace existing workflows right away?

Not automatically. GPT-5.4 looks most useful where a team already mixes documents, spreadsheet models, code changes, and tool use in one flow, so the safer move is to benchmark correction time, token spend, and governance needs before a broad rollout. That recommendation is an inference from OpenAI's workflow-first launch across ChatGPT, Codex, Excel, and the API.

OpenAI Introduces GPT-5.4 for Documents, Spreadsheets, and Code

Updated on March 08, 2026 5 min read

On 5 March 2026, OpenAI introduced GPT-5.4 across ChatGPT, the API, and Codex. The release pushes the GPT-5 line further into everyday professional work, with a stronger focus on documents, spreadsheets, presentations, code, and tool-based tasks.

For Canadian readers, the practical angle is clear. OpenAI also launched ChatGPT for Excel in beta on 5 March 2026 and said rollout had started in Canada, giving the GPT-5.4 launch an immediate spreadsheet workflow tie-in.

The bigger story is not just model quality. OpenAI is packaging reasoning, coding, computer use, and productivity features into a single workflow layer that is meant to sit inside tools people already use.

What happened

OpenAI announced GPT-5.4 on 5 March 2026 and described it as its most capable and efficient frontier model for professional work. The company said GPT-5.4 is available in ChatGPT as GPT-5.4 Thinking, in the API as gpt-5.4, and in Codex for coding and agentic workflows.

The rollout in ChatGPT started with Plus, Team, and Pro users. OpenAI also said Enterprise and Edu admins can enable early access, while GPT-5.4 Pro is available for Pro and Enterprise plans. GPT-5.2 Thinking remains in the Legacy Models section for paid users until 5 June 2026.

OpenAI says GPT-5.4 is the first mainline reasoning model in the GPT-5 family to incorporate the coding capabilities of GPT-5.3-Codex. In Codex, the model is available in the app, CLI, IDE extension, and Codex Cloud, with experimental support for a 1M context window. OpenAI says requests above the standard 272K context window count against usage limits at 2x the normal rate.

A separate OpenAI announcement on 5 March 2026 brought the launch closer to business users. ChatGPT for Excel entered beta the same day, powered by GPT-5.4, with rollout for Business, Enterprise, Edu, Teachers, Pro, and Plus users in the United States, Canada, and Australia. OpenAI also said ChatGPT for Google Sheets is coming soon.

In ChatGPT itself, OpenAI says GPT-5.4 Thinking can show an upfront plan for its reasoning so users can redirect the task while it is in progress. The same release notes highlight stronger work on spreadsheets, presentations, documents, and deep web research.

Why it matters

For learners, GPT-5.4 matters because it aims at the formats that often slow people down in real coursework and junior roles. Writing a clean document, updating a spreadsheet model, tracing assumptions, or turning code changes into a usable output usually takes more than raw technical knowledge. Tools that help across all of those surfaces can reduce friction, especially for people still building confidence.

For developers, the release is more interesting than a benchmark bump. OpenAI is moving toward a single model that can reason, browse, write code, use tools, and work across software environments. That matters if your team has been stitching together different models or assistants for docs, code review, research, and spreadsheet-heavy tasks.

For teams, GPT-5.4 looks less like a standalone chatbot upgrade and more like a workflow product. The Excel add-in, Codex integration, and tool search improvements all point in the same direction: OpenAI wants the model to sit inside everyday work rather than next to it.

That does not remove the need for review. Spreadsheet outputs still need checking, formula logic still needs auditing, and polished code still needs testing. The release makes AI assistance broader, but it does not make human verification optional.

Key numbers

GPT-5.4 scored 83.0% on GDPval, versus 70.9% for GPT-5.2.

On OpenAI's internal investment banking modelling tasks, GPT-5.4 reached 87.3%, versus 68.4% for GPT-5.2.

Human raters preferred GPT-5.4 presentations 68.0% of the time over GPT-5.2 on a presentation prompt set.

OpenAI says GPT-5.4's individual claims are 33% less likely to be false than GPT-5.2, and full responses are 18% less likely to contain any errors on a de-identified set of flagged prompts.

On SWE-Bench Pro, GPT-5.4 scored 57.7%, versus 56.8% for GPT-5.3-Codex and 55.6% for GPT-5.2.

On OSWorld-Verified, GPT-5.4 reached 75.0%, ahead of GPT-5.2 at 47.3% and above the human baseline OpenAI lists at 72.4%.

In a test using 250 tasks from Scale's MCP Atlas with all 36 MCP servers enabled, OpenAI says tool search reduced total token usage by 47% while keeping the same accuracy.

API pricing for gpt-5.4 is listed at 2.50 USD per million input tokens, 0.25 USD per million cached input tokens, and 15 USD per million output tokens.

API pricing for gpt-5.4-pro is listed at 30 USD per million input tokens and 180 USD per million output tokens.

In Codex, GPT-5.4 includes experimental support for a 1M context window.

Context

GPT-5.4 follows GPT-5.3-Codex, which OpenAI introduced on 5 February 2026 as a more capable agentic coding model. The new release matters because OpenAI is no longer treating advanced coding and general professional reasoning as separate tracks. Instead, it is pulling them into one mainline model and reducing the number of choices teams have to make.

That reflects a wider shift in the AI tools market. Buyers are comparing not just writing quality, but also workflow coverage, governance controls, cost per task, and how well a model works inside spreadsheets, IDEs, document flows, and internal tools. GPT-5.4 lands squarely in that broader race.

What's next

The short-term next step is rollout and testing. OpenAI is already connecting GPT-5.4 to ChatGPT, Codex, the API, and the Excel add-in, and it has signalled that Google Sheets support is on the way.

For teams, the sensible move is a controlled pilot. Pick one document-heavy workflow, one spreadsheet task, and one coding task, then measure correction time, token cost, and how often a human still needs to step in. That will tell you more than promotional benchmark tables.

For learners, GPT-5.4 is most useful as a draft-and-debug partner. Use it to outline a report, explain a workbook, or propose code changes, then rebuild the reasoning yourself in Python, Excel, or JavaScript. That is where the skill gain happens.

How to go deeper

Build stronger data and evaluation skills in Code Labs Academy's Data Science & AI Bootcamp.
Strengthen application fundamentals in Code Labs Academy's Web Development Bootcamp.
Explore Code Labs Academy's guide to top workplace AI tool categories to learn in 2026.