Enterprise AI spending is undergoing a massive reality check. As corporate boards demand clear returns on expensive software investments, Anthropic is pivoting its product strategy toward daily efficiency with the official release of Claude Opus 5.
Positioned right below its top-tier Fable 5, the new AI model aims to deliver near-frontier intelligence at half the operational cost. Anthropic kept API pricing at $5 per million input tokens and $25 per million output tokens—the same rate as its predecessor, Opus 4.8. The model immediately assumes the role of default driver across Anthropic’s flagship products, taking top billing on the consumer Claude Max plan and securing the primary spot on Claude Pro.
High benchmark scores with fewer tokens
Rather than chasing raw capability records at any financial expense, Anthropic engineered Opus 5 to handle bounded professional tasks with maximum token efficiency. On terminal coding evaluation Frontier-Bench v0.1, Opus 5 notched a 43.3% score. These figures surpass both its predecessor and Fable 5 while consuming fewer computational resources.
Early-access enterprise partners reported immediate cost benefits across specialized industries. AI legal provider Harvey reported that Opus 5 matched maximum-reasoning outputs while generating 26% fewer tokens on average, while Zapier Chief Executive Wade Foster noted that the model completed a complex churn-prevention sequence end-to-end on AutomationBench—a test previous models failed entirely (via Quartz). Meanwhile, coding assistant platform Cursor confirmed that Opus 5 delivers near-Fable intelligence at significantly higher execution speeds (via ZDNet).
To give developers finer control over API expenses, Anthropic also added a built-in effort toggle. This allows teams to intentionally reduce reasoning depth on simpler prompts, saving tokens and speeding up response times.
Refined guardrails and flexible tool routing
Safety and reliability also received practical upgrades. Anthropic reduced trigger sensitivity on safety classifiers by roughly 85% compared to Fable 5. This addresses widespread developer frustration over overly rigid refusal triggers during legitimate cybersecurity testing.
When a safety classifier does intervene, API requests automatically route to secondary models rather than returning dead-end error messages. Furthermore, a new mid-conversation tool swapping feature lets developers remove external tools dynamically without wiping the prompt cache, preventing unnecessary token burn.
The post Anthropic Launches Claude Opus 5: Smarter, Faster and Half the Price appeared first on Android Headlines.
​Â
