Anthropic releases Claude Opus 4.8: Check what’s new

Anthropic has accelerated its upgrade cadence once again. Claude Opus 4.8, released today, arrives just six weeks after version 4.7, signalling that the company is ramping up the pace of model improvements.

Agentic coding scores jump from 64.3% to 69.2%, a 4.9% point improvement. Multidisciplinary reasoning with tools improves from 54.7% to 57.9%. Knowledge work scores rise from 1753 to 1890. Agentic computer use gains 0.6 percentage points, moving from 82.8% to 83.4%. Financial analysis improvements reach 53.9% from 51.5%.

These incremental but consistent gains across multiple domains suggest that Anthropic's model training approach is delivering results systematically, rather than focusing on single metrics.

The Fast mode sees considerable optimisation. Anthropic says that Claude Opus 4.8 fast mode is 2.5x faster compared to its predecessor but costs three times less.

This double optimisation, which is faster and cheaper at the same time, solves one of the primary pain points for both developers and organisations dealing with heavy workloads. The pricing on Opus 4.8 in general does not differ from Opus 4.7 in any way.

But aside from the model update, Anthropic has another three updates coming up. Dynamic Workflows, currently under research preview, will enable Claude to deal with more extensive tasks on Claude Code. Effort Control in Claude.ai and Cowork will allow you to define how much computing power will be used by Claude to give you an answer.