Anthropic has unveiled Claude Opus 5, its new flagship model for coding, agents and enterprise workflows – at roughly half the price of its previous top model Fable 5, and with a new “effort” dial.
The effort dial
The most striking new feature is an adjustable “effort” setting. It lets you deliberately trade off maximum intelligence against lower cost: for your highest-value tasks the model runs at full power, while for simpler tasks it runs cheaper and faster. Anthropic’s charts explicitly emphasise performance per cost, not just peak performance.
Coding and agents
On the agentic terminal benchmark Frontier-Bench v0.1, Opus 5 scores 43.3 percent according to Anthropic – more than double Opus 4.8 (18.7 percent) and well ahead of Fable 5 (33.7 percent), at a lower cost per task. Across effort levels the model averaged 9 percentage points higher accuracy – with about a third fewer steps and tool calls and 60 percent less time. Opus 5 was released on 24 July 2026.
Sources: Anthropic, VentureBeat.
