🏢 Enterprise Adoption Hits Milestone
OpenAI disclosed on May 14, 2026, that GPT-5 is now deployed in customer-facing service roles across 40% of Fortune 500 companies, up from 21% in January 2026. Major deployments include JPMorgan Chase (retail banking chat and email routing), UnitedHealth Group (claims explanation and prior authorization), Walmart (in-store associate assistant via handheld devices), and Salesforce (Einstein GPT agent for CRM workflows).
OpenAI's enterprise revenue reached $4.2 billion in Q1 2026, representing 72% of total revenue and reflecting a strategic pivot from consumer subscriptions to enterprise contracts.
GPT-5 Turbo—a latency-optimized variant—delivers 98.3% reduction in response time for common enterprise queries compared to GPT-4, with p99 latency under 80ms for tasks like classification, extraction, and summarization. This has unlocked real-time use cases including live-call transcription and response suggestion where earlier models introduced unacceptable delays. The enterprise API pricing has dropped 60% year-over-year to $1.75 per million input tokens.
đź“‹ o3 Reasoning and Sora Video Integration
OpenAI introduced the o3 reasoning endpoint in May 2026, offering chain-of-thought reasoning with adjustable "thinking budgets" from 10 seconds to 10 minutes. In benchmarking, o3 with a 10-minute budget scores 96.7 on the 2024 AIME mathematics competition and 87.3 on GPQA Diamond (graduate-level science questions), achieving performance levels that exceed human PhD-level specialists in chemistry and physics.
The endpoint is priced at $30 per million output tokens to account for the hidden reasoning tokens that significantly exceed visible output.
The GPT-5 multimodal endpoint now integrates Sora, OpenAI's video generation model, enabling text-to-video generation natively through the chat API. Users can request video content alongside text analysis—for example, asking GPT-5 to analyze a financial report and generate a 30-second animated summary video. Pricing for video generation is $0.025 per second of generated video, placing a 60-second clip at $1.50.
OpenAI reports early adoption by marketing agencies and e-learning platforms.
đź“‹ Operator Agent Framework Goes GA
The Operator agent framework, which allows GPT-5 to control a web browser to complete multi-step tasks, graduated from research preview to general availability. Operator can navigate websites, fill forms, extract data across multiple pages, and integrate results into structured outputs. OpenAI reports a 91.2% task completion rate on WebArena, a benchmark of real-world web tasks, with most failures attributable to CAPTCHA challenges and dynamic JavaScript rendering issues.
Operator is available only to enterprise customers and requires explicit per-domain opt-in for security reasons. The framework has been particularly popular in e-commerce (price comparison and purchasing automation), travel booking, and financial document retrieval.