Antropic Unveils Opus 5: Surprising Performance, Outpricing Fable 5 by Half

Antropic has officially launched Opus 5, the anticipated successor to its Opus 4.8 model, delivering a notable surprise to the AI community. Despite initial expectations of being less powerful than Fable 5, Opus 5 has demonstrated superior performance across numerous benchmarks, crucially doing so at half the price. This new model is set to become the default in Antropic’s ecosystem tools, including Claude Code, providing developers with Fable 5-level capabilities without the prohibitive token expenditure. Specifically, Opus 5’s pricing is set at $5 per million input tokens and $25 per million output tokens, a direct reduction from Fable 5’s $10 input and $50 output per million tokens, making it a highly attractive option for cost-conscious development.

Opus 5’s benchmark victories are particularly pronounced in areas critical for software development: coding tasks, agentic operations, and ‘computer use’ scenarios involving program control and manipulation. Beyond these, the model shows significant advancements in visual capabilities, adept at generating graphics, animations, and detailed explanatory images, including 3D animations with animated or real-world objects. However, this focused improvement comes with deliberate trade-offs; Opus 5 does not show similar enhancements in legal or health-related topics, where it reportedly performs worse than Fable 5. This strategic limitation underpins Antropic’s intent to position Opus 5 as a general-purpose, high-performance coding and visual model, while intentionally not training it on cyberattack tasks, making it inherently more inoffensive than counterparts like Mito 5 or Fable 5. This also differentiates it from Mito 5, which remains Antropic’s more cybersecurity-focused offering.