OpenAI has rolled out GPT-5.6 Sol with Ultrafast mode, a new API service tier delivering up to 750 output tokens per second—fourteen times faster than standard inference. Built on Cerebras hardware, the tier addresses a critical pain point for enterprises: latency-sensitive workloads that require real-time responsiveness. Real-time customer support automation, live code generation, and instant data analysis now become viable at scale. The speed matters because inference latency directly impacts user experience in agentic workflows, where AI agents autonomously handle multi-step tasks. Competitors like Anthropic and smaller inference-optimization startups have been gaining traction with speed claims, but OpenAI's integration of Cerebras and native API support gives it a significant edge in production environments where enterprises demand both capability and performance guarantees.
To capitalize on this technical advantage, OpenAI appointed Dali Rajic as Chief Revenue Officer. Rajic, formerly a senior executive at enterprise software firms, brings deep experience in converting platform capabilities into repeatable enterprise revenue models. His hire signals OpenAI's shift from product-first to sales-first leadership—a critical signal that the company is prioritizing commercial execution over raw model scaling. OpenAI's enterprise revenue already surpassed consumer revenue for the first time, reaching $40 billion ARR two quarters ahead of internal targets. RingCentral exemplifies the ROI: the communications platform used ChatGPT Work and Codex to centralize operational intelligence across engineering and operations, accelerating product development cycles. A second customer example, though unnamed in recent releases, shows similar patterns where enterprises using the Responses API for multi-turn agent workflows achieved cost reductions of 30-40% through smarter model selection and routing.
The timing is urgent. Microsoft and Google are embedding AI agents into enterprise workflows through their own investments in inference infrastructure and sales organizations. Anthropic is targeting enterprise security concerns with its Constitutional AI model claims. If OpenAI stumbles on execution—missing SLA targets, overpricing the Ultrafast tier, or failing to provide enterprise-grade support through Rajic's organization—it risks ceding wallet share in the most profitable segment of the AI market. The next 18 months will reveal whether OpenAI can convert technical leadership into durable enterprise lock-in. The speed tier and CRO hire suggest they're taking that fight seriously.