GPT-5.4 closes the context window gap with Claude Opus, both now at 1M tokens. The more significant number is the 83% GDPVal score - a benchmark measuring end-to-end completion of real economic tasks (emails, reports, spreadsheet work, analysis). An 83% score represents a roughly 8-point jump over GPT-5.1 and puts the model within reach of expert human performance on knowledge-work tasks.
OpenAI also announced ARR surpassing $25B in April 2026, with the company reportedly taking early steps toward an IPO as soon as late 2026. The combination of strong model performance and revenue momentum positions OpenAI for a significant public offering.
The context-window race is effectively over - both frontier models now handle 1M tokens. Evaluate GPT-5.4 vs Claude Opus 4.7 on your specific workflow, not benchmarks. Most founders will use both.
