OpenAI Launches GPT-5.5 Reasoning Tier With Hourly-Compute Pricing For Enterprise Workloads

OpenAI has launched GPT-5.5, an expanded reasoning-tier variant of its flagship foundation model family, alongside an unconventional hourly-compute pricing structure for enterprise workloads that represents a meaningful departure from the per-token pricing convention that has anchored the foundation-model commercial-pricing landscape since the GPT-3.5 era.

The GPT-5.5 reasoning tier is positioned as a deeper-reasoning extension of the GPT-5 base model, with substantially expanded chain-of-thought reasoning depth that delivers measurably better performance on complex multi-step reasoning tasks at the cost of materially-higher per-query latency and compute consumption. The benchmark performance places GPT-5.5 ahead of GPT-5 by roughly 30-45% on the principal scientific-reasoning, advanced-mathematics, and software-engineering evaluation suites, with the largest gains concentrated in the categories where multi-step deliberation produces the highest marginal-quality improvement.

The hourly-compute pricing structure is the more strategically interesting commercial framing. Enterprise customers can purchase committed compute capacity at a flat per-hour rate against a designated isolated-cluster of GPU resources, rather than paying per-token for individual queries. The pricing is structured at $50 per H300-equivalent hour with substantial volume discounts for multi-month commitments, and the framework is positioned as an alternative for workloads where the per-token pricing creates either unpredictability or unfavourable unit-economics relative to the intrinsic compute cost.

The strategic logic behind the hourly-pricing model is substantially about competitive positioning against the major hyperscalers' own first-party AI offerings. AWS Bedrock, Azure OpenAI Service, and Google Cloud Vertex AI all offer enterprise-tier capacity reservations against the underlying compute infrastructure, but the pricing-and-commercial-framework has been comparatively-less-flexible than what enterprise customers have been increasingly demanding. OpenAI's hourly-compute offering is a direct response to that customer-side pressure and is positioned to substantially reduce the friction for the largest enterprise deployments.

For the wider foundation-model commercial-pricing landscape, the GPT-5.5 hourly-compute structure may represent a meaningful early signal of structural change. The per-token pricing convention has been increasingly strained by the growing share of enterprise workloads where the variable-cost-per-query economics are difficult for both the frontier-lab supplier and the enterprise customer to manage. Whether the hourly-compute model becomes a broader industry convention through the rest of the year — and whether Anthropic, Google DeepMind, and the wider commercial-AI complex follow OpenAI's pricing-architecture lead — is the principal forward variable to track.