Anthropic Opus 5 Gives Frontier Performance at Half the Price

Anthropic’s Claude Opus 5 delivers near-frontier performance at roughly half the price, which is a major strategic win for them. This puts competitive pressure on OpenAI, and a clear advance of the practical frontier toward cost-efficient agentic intelligence.

It sets new standards for agentic coding, knowledge work, science and business workflows.

Metrics overview

Opus 5 leads or is highly competitive across key agentic and professional tasks.

Strongest leads
Agentic terminal coding (Frontier-Bench v0.1) at 43.3% vs 33-34% for other near frontier. Even better than their own Fable 5.
Novel problem-solving (ARC-AGI-3) at 30.2% (far ahead of others)
Knowledge work (GDPval-AA v2) at 1861
Computer use (OSWorld 2.0) at 70.6%
Business workflows (AutomationBench) at 26.0%
Hard biology at 49.4%.

Very competitive: Agentic search (90.8%), multidisciplinary reasoning (especially with tools at 64.7%), and FrontierCode agentic coding (~53.4%, essentially tied with Fable 5).

Trailing in spots: Slightly behind on some pure coding DeepSWE), legal, and health benchmarks (where Fable/Mythos or GPT edge it).

XAI Grok 4.6 will be out in August and then Grok 5 in September or October.