The new flagship model rivals Anthropic's own Fable 5 on coding, reasoning, and agentic tasks while keeping the same pricing as Opus 4.8.

Artificial intelligence models are getting smarter. But for most developers and businesses, they're also getting more expensive.

Anthropic is betting that doesn't have to be the case.

The company has introduced Claude Opus 5, its latest flagship model, positioning it as a high-performance alternative that comes close to the capabilities of Claude Fable 5 while costing roughly half as much. Available today across Claude, Claude Code, and the Anthropic API, Opus 5 targets developers, researchers, and enterprises that need frontier-level intelligence without paying premium model prices.

Instead of chasing an entirely new tier of intelligence, Anthropic's latest release focuses on making its strongest capabilities more accessible.

Anthropic says Claude Opus 5 excels across the tasks businesses increasingly rely on AI for:

According to the company's published benchmarks, Opus 5 outperforms both Claude Fable 5 and OpenAI's GPT-5.6 Sol on several coding and agentic evaluations while maintaining the same API pricing as Claude Opus 4.8.

One of the biggest highlights is its performance on ARC-AGI-3, a benchmark designed to measure reasoning beyond memorization.

Anthropic reports that Opus 5 achieved 30.2%, roughly three times higher than the next-best published result. The model also moved to the top of Artificial Analysis' Intelligence Index leaderboard, an independent ranking that compares leading foundation models across reasoning and knowledge tasks.

Anthropic shared several results demonstrating where Opus 5 improves over previous models.

Perhaps the most surprising result came from the International Math Olympiad 2026, where Anthropic says Opus 5 solved all 42 problems, comfortably exceeding the 29-point threshold typically associated with a gold medal.

While benchmark scores attract headlines, pricing may ultimately matter more for many organizations.

Claude Opus 5 costs:

Those prices are unchanged from Claude Opus 4.8, making Opus 5 significantly less expensive than Claude Fable 5 while delivering comparable performance on many practical workloads.

For companies running AI-powered coding assistants, internal knowledge systems, or enterprise workflows, lower inference costs can translate into substantial savings at scale.

Raw capability wasn't the only focus of this release.

Anthropic describes Opus 5 as its most aligned model to date, claiming it performs similarly to Claude Mythos 5 at identifying software vulnerabilities while remaining considerably less capable of generating offensive cyber exploits.

That distinction reflects Anthropic's broader strategy of improving capability while reducing opportunities for misuse, an area that has received increased regulatory attention as frontier AI systems become more powerful.

Over the past year, AI companies have increasingly segmented their model lineups.

Instead of one flagship model for every task, providers now offer multiple options optimized for different combinations of performance, speed, and cost.

Claude Opus 5 fits squarely into that strategy.

Rather than replacing Fable 5, Anthropic is offering developers a model that delivers much of the same intelligence for everyday software engineering, research, and enterprise work at a far more accessible price point.

For businesses evaluating AI infrastructure costs, that balance could prove more valuable than incremental benchmark gains alone.

Claude Opus 5 is available immediately through the Claude web app, Claude Code, and the Anthropic API. The release also continues Anthropic's rapid rollout of new models, with four major Claude 5-family launches in less than two months.

As OpenAI, Google, and Anthropic continue competing on intelligence, speed, and cost, the next phase of the AI race may be less about building the most powerful model and more about delivering the best value for developers and enterprises.