Anthropic Releases Claude Opus 5 AI Model with Near-Flagship Performance at Half the Cost

Jul 24, 2026

Anthropic has launched Claude Opus 5, a new AI model designed for daily enterprise use that delivers near-flagship Fable 5 performance at half the cost. Priced identically to its predecessor, Opus 4.8, at $5 per million input tokens, the model introduces adjustable effort settings to help businesses manage token expenses. Opus 5 demonstrates superior self-verifying capabilities and sets new benchmarks in coding and biology, while maintaining strict safety guardrails that restrict offensive cyber exploitation.

Claude Opus 5 release

▪Claude Opus 5 is the new default model on Claude Max and the strongest model available on Claude Pro.
▪Anthropic introduced a Fast mode for Claude Opus 5 that runs at 2.5 times the default speed for twice the base price.
▪Claude Opus 5 is priced at $5 per million input tokens and $25 per million output tokens, which is the same price as Opus 4.8.
▪Anthropic released Claude Opus 5 on July 24, 2026, as an upgrade to its Opus 4.8 model.

Benchmark performance comparisons

▪On an internal organic chemistry benchmark, Claude Opus 5 scored 10.2 percentage points higher than Opus 4.8 at inferring molecular structures from spectroscopy data.
▪Claude Opus 5 scored three times higher than the next-best model on the ARC-AGI 3 novel problem-solving evaluation.
▪Claude Opus 5 scored 43.3% on the Frontier-Bench v0.1 agentic terminal coding benchmark, compared to 18.7% for Opus 4.8 and 33.7% for Fable 5.
▪On the OSWorld 2.0 computer-use benchmark, Claude Opus 5 surpassed Fable 5's best result at just over one-third of the cost.

Token efficiency economics

▪Claude Opus 5 features an adjustable effort setting that allows users to toggle how much effort the model expends to balance cost and capability.
▪Harvey reported that Claude Opus 5 achieved similar performance to Opus 4.8's maximum-reasoning mode while generating 26% fewer tokens on average.
▪Fundamental Research Lab reported that Claude Opus 5 averaged 9 percentage points higher accuracy on financial-modeling tasks while using one-third fewer turns and 60% less time.

Self-verifying agent capabilities

▪Stripe staff software engineer Cristian Rivera reported that Claude Opus 5 successfully managed development environments autonomously over a weekend, building its own monitor and driving each box.
▪Claude Opus 5 is designed to verify its own work and iterate autonomously, such as writing its own computer vision pipeline to extract geometry from raw pixels during a FreeCAD reconstruction task.

Safety controls capability gaps

▪Anthropic intentionally avoided training Claude Opus 5 on cyber tasks, leaving it substantially behind Mythos 5 on vulnerability exploitation despite coming close on vulnerability identification.
▪Claude Opus 5 scored 2.3 on Anthropic's automated behavioral audit for overall misaligned behavior, which is lower than Opus 4.8, Sonnet 5, or Fable 5.
▪When a safety classifier declines a request in Claude Opus 5, the system can automatically fall back to Opus 4.8.
▪Claude Opus 5's cyber classifiers block binary-based vulnerability scanning, penetration testing, and exploit generation, but are expected to intervene 85% less often than those on Fable 5.

Anthropic valuation funding

▪Anthropic has secured a reported $30 billion Azure compute deal alongside infrastructure arrangements with Google Cloud and Nvidia.
▪A U.S. judge granted final approval to Anthropic's $1.5 billion copyright settlement with book authors in July 2026.
▪Reuters reported in February 2026 that Anthropic was valued at approximately $380 billion in its latest funding round.