Introducing GPT‑5.4 — news
News/2026-03-08-introducing-gpt54-news-news
Breaking NewsMar 8, 20264 min read

Introducing GPT‑5.4 — news

Featured:OpenAI
Introducing GPT‑5.4 — news

OpenAI Launches GPT-5.4 with Enhanced Professional Capabilities and 1M Token Context

OpenAI announced GPT-5.4 on March 5, 2026, introducing two new API models — gpt-5.4 and gpt-5.4-pro — that are also available in ChatGPT and Codex CLI. The models feature an August 31, 2025 knowledge cutoff and a 1 million token context window, with pricing set slightly higher than the GPT-5.2 family and additional costs for inputs exceeding 272,000 tokens. OpenAI positioned the release as its most capable and efficient frontier model for professional work, showing particular gains in business applications such as spreadsheet modeling.

OpenAI released GPT-5.4 as it seeks to maintain its lead in the frontier AI race amid growing competition from Anthropic’s Claude models, which have recently emphasized enterprise use cases. The new models build directly on the GPT-5.2 family with improvements in reasoning, coding, and multimodal understanding.

According to OpenAI’s official announcement, GPT-5.4 demonstrates strong performance on professional tasks. On an internal benchmark of spreadsheet modeling tasks similar to those performed by junior investment banking analysts, GPT-5.4 achieved a mean score of 87.3%, compared to 68.4% for GPT-5.2.

The company also highlighted improvements in creating and editing spreadsheets, presentations, and documents. GPT-5.4 beats the coding-specialist GPT-5.3-Codex on relevant benchmarks, prompting speculation about whether the Codex model line will be merged into the main GPT-5.4 family.

Technical Specifications and Availability

Both gpt-5.4 and gpt-5.4-pro support a 1 million token context window, representing a significant expansion in long-context capabilities. The models include enhanced image input processing: starting with GPT-5.4, OpenAI introduced an original image input detail level that supports full-fidelity perception up to 10.24 million total pixels or a 6000-pixel maximum dimension, whichever is lower. The high image input detail level now supports up to 2.56 million total pixels or a 2048-pixel maximum dimension.

The models are rolling out starting today and are available in ChatGPT, Codex, and OpenAI’s API. GPT-5.4 Thinking, a reasoning-focused variant, is available for Plus, Teams, and Pro users. GPT-5.4 Pro is accessible through the API and for select ChatGPT users. The models are also being integrated into Microsoft Foundry, where Microsoft described GPT-5.4 as “a model designed to help organizations move from planning work to reliably completing it in production environments.”

Performance and Demonstrations

OpenAI and independent testers have shared examples of GPT-5.4’s multimodal generation capabilities. Simon Willison, a prominent AI developer and blogger, published examples of cartoon pelicans riding bicycles generated by both GPT-5.4 and the more powerful GPT-5.4 Pro variant. The Pro version took 4 minutes and 45 seconds to generate one image and cost $1.55 according to LLM pricing data.

TechCrunch reported that OpenAI is also releasing GPT-5.4 with Pro and Thinking versions, positioning the model as particularly suited for professional workflows. Gizmodo noted the release comes at a time when OpenAI appears to be seeking a significant market win.

Impact on Developers and Enterprise Users

The improvements in spreadsheet and document handling could prove valuable for financial services, consulting, and other data-heavy industries. The jump from 68.4% to 87.3% on the internal investment banking benchmark suggests meaningful progress toward automating complex analytical tasks that previously required substantial human oversight.

For developers, the 1 million token context window opens new possibilities for building applications that can process and reason over extremely long documents, codebases, or conversation histories. The pricing structure, which increases above 272,000 tokens, encourages efficient prompt engineering while still allowing access to the full context when necessary.

The enhanced image understanding capabilities further strengthen GPT-5.4’s position in multimodal applications, potentially accelerating adoption in design, analysis, and content creation tools that combine text and visual reasoning.

What's Next

OpenAI has not yet detailed a specific timeline for additional GPT-5.4 variants or the next major model release. Industry observers will be watching whether the company continues the incremental .x versioning approach or shifts to a new major version number in the coming months.

Integration into Microsoft Foundry suggests deeper enterprise deployment is already underway. Developers can begin experimenting with the models immediately through the OpenAI API and ChatGPT interfaces.

The release continues the rapid iteration pattern seen in the GPT-5 series, with OpenAI delivering capability improvements roughly every few months while expanding context windows and multimodal features.

Sources

Original Source

simonwillison.net

Comments

No comments yet. Be the first to share your thoughts!