Key Highlights
- GPT-5.4 became available March 5, 2026, representing OpenAI’s strongest professional-focused model to date
- Financial professionals gain access to integrated FactSet and Third Bridge data platforms
- Direct integration works within Microsoft Excel and Google Sheets environments
- The release positions OpenAI alongside Anthropic’s Claude for Financial Services in the enterprise market
- ChatGPT Plus, Team, and Pro members receive immediate access to the model
OpenAI introduced GPT-5.4 to the market on Thursday, March 5, 2026. The company focused this iteration on capabilities that serve finance sector professionals.
This AI system creates spreadsheets, documents, and slide presentations while minimizing the need for repeated user prompts. Web search functionality allows the model to collect information and address sophisticated inquiries.
The company embedded specialized financial capabilities that link to FactSet Research Systems and Third Bridge data sources. These connections target professionals performing market analysis and creating investment documentation.
Direct model access now works within [[LINK_START_0]]Microsoft[[LINK_END_0]] Excel and Google Sheets applications. OpenAI simultaneously released a ChatGPT add-in compatible with Excel.
This launch intensifies competition with Anthropic, whose Claude for Financial Services entered the market previously. Both organizations pursue corporate clients seeking premium AI solutions.
Model Performance Metrics
Internal testing focused on investment banking spreadsheet operations showed GPT-5.4 achieving 87.3%, compared to 68.4% from its GPT-5.2 predecessor. Human evaluators selected GPT-5.4 presentations 68% of the time when comparing outputs against GPT-5.2.
GDPval assessment, which evaluates AI performance across 44 professional fields, showed GPT-5.4 equaling or surpassing human experts in 83% of test scenarios. GPT-5.2 achieved 70.9% on identical evaluations.
The model attained 75% on OSWorld-Verified, an assessment measuring desktop navigation abilities using visual input and cursor control. Human benchmarks for this evaluation stand at 72.4%.
OpenAI characterizes this release as delivering superior factual accuracy. Individual assertions show 33% fewer errors relative to GPT-5.2 outputs.
Access Tiers and Cost Structure
GPT-5.4 deployment begins today for ChatGPT Plus, Team, and Pro members under the designation GPT-5.4 Thinking. GPT-5.2 Thinking remains operational through a three-month transition period ending June 5, 2026.
Enterprise and Edu account administrators can activate early access through platform settings. GPT-5.4 Pro serves Pro and Enterprise subscription tiers.
API access carries rates of $2.50 per million input tokens and $15 per million output tokens. These figures exceed GPT-5.2 rates of $1.75 and $14 respectively.
Context window support extends to one million tokens within Codex. Standard API calls handle 272,000 tokens, while larger requests incur double standard rates.
OpenAI emphasizes that GPT-5.4 delivers improved token efficiency versus GPT-5.2, offsetting elevated per-token costs through reduced overall usage.
The company makes the model accessible through Codex, its AI-powered coding platform, and via standard API access using the identifier gpt-5.4.
