Anthropic on July 24, 2026, released Claude Opus 5, its newest flagship, carrying the same sticker price as the model it replaced while asserting materially stronger performance. Opus 5 is priced at $5 per million input tokens and $25 per million output tokens, the same rate as its predecessor Opus 4.8, and Anthropic says it approaches the capabilities of its top-tier Claude Fable 5 model across many tasks. Fable 5, launched in June, is priced at $10 per million input tokens and $50 per million output tokens, making Opus 5's cost parity with the older model a pointed value argument. Opus 5 is Anthropic's fourth Claude 5 model release in less than two months, underscoring how AI deployment has shifted from blockbuster launches to rapid improvements on capability, cost, and speed.
The architecture's most consequential new feature is an effort dial that directly addresses a complaint Fable 5 generated at enterprise scale. Opus 5 comes with a feature enabling users to toggle how much effort, low, medium, or high, the model expends completing a task or answering a prompt. Business customers and developers had criticized Fable 5's burn rate, the high number of tokens the model used when completing tasks, which resulted in users blowing through their allotted token budgets or running up large bills. Anthropic's own benchmark figures give the toggle quantitative weight: on Frontier-Bench v0.1, Opus 5 surpasses all other models and more than doubles Opus 4.8's performance at a lower cost per task, and on CursorBench 3.2 at maximum effort it lands within 0.5 percent of Fable 5's peak score at half the cost per task. Real-world confirmation came from Harvey, the legal-AI firm, whose head of applied research reported that Opus 5 matched the output quality of Opus 4.8 running at maximum reasoning while cutting average token usage by 26 percent. The model ships with a one-million-token context window, is the new default on Claude Max, and carries no data-retention requirements for general access, unlike Anthropic's two most capable models, a compliance detail that may matter as much as benchmark scores to regulated buyers.
The competitive timing is hard to ignore. OpenAI said Friday it has suspended work on some aspects of its upcoming model Astra after an internal review found it had made significant advancements in agentic coding and cybersecurity, enough to warrant concern over its capabilities. That model reached OpenAI's "critical cybersecurity threshold," meaning it could independently identify and carry out cyberattacks against traditionally well-protected real-world systems. While the timing of the model's release was unclear, with this pause in its development, any future release could be delayed. The gap leaves Anthropic with a pricing and feature argument that faces no imminent generational counter-punch from its largest rival.
Skeptics, however, note that the performance claims rest on a narrow evidential base. The benchmark figures are Anthropic's own evaluations, run on Anthropic's own harnesses; they are self-reported, and independent replication takes weeks, meaning every figure should be treated as a vendor claim rather than settled fact. Practitioner feedback has been mixed: a large number of developers and knowledge workers report that Opus 5 feels worse than its predecessor Opus 4.8 in daily work, especially coding. The system card adds its own friction: reported problems include overconfidence on unsupported claims and overdramatic phrasing, superlatives, and apologies, a pattern critics call an ongoing Claude family trait. Notably, Anthropic's system card reports the model estimates a 41 percent likelihood that it qualifies as a moral patient, a meaningful increase from prior Claude models, a disclosure the company chose to publish but that independent researchers have no current means to verify.
The practical question for any team with an AI bill is blunt: Opus 5 is cheaper than Fable 5 and, on Anthropic's own numbers, nearly as capable, but until independent labs replicate those benchmarks, the word 'nearly' is doing a lot of work.