Anthropic launches Claude Opus 5.5, cuts costs 40%
Anthropic unveiled Claude Opus 5.5 on September 22, 2026, marking the first release in its new Claude 5.5 family and cementing the model as the company's most powerful option for long-running agent work, according to Reuters.
The model will cost $4 per million input tokens and $20 per million output tokens, pricing that Reuters reported reflects a 40% reduction in compute cost compared to Opus 5 on typical workloads. The pricing structure positions Opus 5.5 as a direct competitor to OpenAI's latest flagship offerings while maintaining Anthropic's emphasis on long-context reasoning—a core requirement for agents managing stateful, multi-step operations.
Immediate Availability Across Cloud Platforms
Unlike typical staged rollouts, Opus 5.5 is available immediately across Amazon Web Services, Google Cloud, and Microsoft Azure, according to Anthropic's official product page. This multi-cloud distribution reduces vendor lock-in friction for enterprise teams already committed to specific infrastructure providers and accelerates adoption among teams building production agents.
The move reflects Anthropic's strategy to embed Claude deeper into enterprise AI stacks. By launching simultaneously on three major cloud providers, the company signals readiness to compete at scale in the agent economy where deployment flexibility matters as much as model capability.
Why the Timing Matters for Agents
Agent applications—software that autonomously executes tasks across APIs, databases, and external systems—require models that can maintain context across extended interactions, reason about complex environments, and handle long chains of thought without degradation. Opus 5.5's positioning as the first release in a new family suggests Anthropic plans sustained investment in this tier, reducing risk for teams building production agent infrastructure on top of Claude.
The cost reduction is material for agent operators. A typical agent might process hundreds of thousands of tokens daily across reasoning, planning, and interaction phases. A 40% cost reduction on compute translates directly to margin improvement for service providers or operational savings for enterprises deploying large agent fleets.
Market Position
Ophir Frieder, head of product at Anthropic, emphasized that Opus 5.5 achieves "fable-level performance"—a reference to internally benchmarked capabilities comparable to Anthropic's most capable models—while operating at lower cost. This is concrete positioning: the company is not claiming new capabilities but rather delivering existing capability at reduced price.
For agent teams evaluating inference engines, the combination of availability, pricing, and multi-cloud support removes friction from evaluation-to-deployment timelines.