TLDR
- OpenAI introduced GPT-5.4, its most advanced professional AI model, on March 5, 2026
- New financial capabilities include integrations with FactSet and Third Bridge data services
- Direct access within Microsoft Excel and Google Sheets now available
- The release intensifies competition with Anthropic’s Claude for Financial Services offering
- Access begins immediately for ChatGPT Plus, Team, and Pro tier subscribers
On Thursday, March 5, 2026, OpenAI unveiled GPT-5.4, its newest artificial intelligence model specifically designed with financial industry professionals in mind.
This advanced model streamlines the creation of spreadsheets, documents, and presentations by requiring fewer conversational iterations. It performs web searches to compile data and deliver comprehensive responses to sophisticated inquiries.
The release features specialized financial capabilities that link directly to FactSet Research Systems and Third Bridge, two major data providers. These connections serve professionals conducting financial analysis and preparing investment documentation.
The model now functions natively within Microsoft Excel and Google Sheets environments. OpenAI simultaneously released a dedicated ChatGPT add-in for Excel alongside this model.
This development escalates competitive pressure on Anthropic, which introduced Claude for Financial Services in the previous year. Both organizations are vying for corporate clients prepared to invest in premium AI solutions.
GPT-5.4 Performance and Benchmarks
When evaluated on OpenAI’s internal investment banking spreadsheet assessment, GPT-5.4 achieved an 87.3% score, representing a significant jump from GPT-5.2’s 68.4% result. Human evaluators favored GPT-5.4-generated presentations 68% of the time when compared against GPT-5.2 output.
The GDPval benchmark, which evaluates AI performance across 44 different professional roles, showed GPT-5.4 matching or exceeding human expert performance in 83% of evaluated scenarios. The earlier GPT-5.2 version achieved 70.9% on this same assessment.
GPT-5.4 recorded a 75% score on OSWorld-Verified, a benchmark that evaluates how effectively an AI system can operate a desktop computer using visual input and cursor control. For context, human participants score an average of 72.4% on this evaluation.
OpenAI characterizes this release as its most accurate model to date. The frequency of factually incorrect statements dropped by 33% compared to GPT-5.2’s baseline.
Availability and Pricing
Starting today, GPT-5.4 becomes accessible to ChatGPT Plus, Team, and Pro subscribers under the designation GPT-5.4 Thinking. The preceding GPT-5.2 Thinking model will remain operational for a three-month transition period before discontinuation on June 5, 2026.
Enterprise and Edu account administrators can activate early access through their management consoles. GPT-5.4 Pro is offered to subscribers on Pro and Enterprise tiers.
API access begins at $2.50 per million input tokens with output tokens priced at $15 per million. These rates exceed GPT-5.2’s pricing structure, which stood at $1.75 and $14 respectively.
The model accommodates up to one million tokens of context when used through Codex. Standard API calls handle 272,000 tokens, while larger requests incur double the standard rate.
According to OpenAI, GPT-5.4’s improved token efficiency offsets the increased per-token cost, ultimately delivering lower overall expenses for typical use cases.
Developers can access the model through OpenAI’s Codex programming assistant or via the standard API using the identifier gpt-5.4.
Get 3 Free Stock Ebooks
Discover top-performing stocks in AI, Crypto, and Technology with expert analysis.
- Top 10 AI Stocks - Leading AI companies
- Top 10 Crypto Stocks - Blockchain leaders
- Top 10 Tech Stocks - Tech giants
