Anthropic just cut the price of frontier intelligence in half

Anthropic released Claude Opus 5 at the same price as Opus 4.8 ($5 in, $25 out), half the cost of Fable 5, while achieving within 0.5% of Fable 5 on CursorBench 3.2 and tripling the next-best model on ARC-AGI 3. The model also ships with 85% fewer safety blocks, 2.5x faster speed in Fast mode, and enables Harvey to match Opus 4.8's max-reasoning quality using 26% fewer tokens.

Anthropic just cut the price of frontier intelligence in half Claude Opus 5 lands within 0.5% of Fable 5 on coding at half the cost, triples the field on ARC-AGI 3, and ships with 85% fewer safety blocks. The full playbook: effort dial, routing table, prompts, m Anthropic shipped Claude Opus 5 today, and the headline number is the one they refused to raise. The price. Same $5 in, $25 out as Opus 4.8. Half of what Fable 5 charges. And the performance sits close enough to the frontier that the tier below it just became the default for almost everyone: ▫️ Frontier-Bench v0.1: 43.3% , against Opus 4.8’s 18.7% and Fable 5’s 33.7% ▫️ CursorBench 3.2: within 0.5% of Fable 5’s peak , at half the cost per task ▫️ ARC-AGI 3: three times the next-best model , on a test built so memorization fails ▫️ OSWorld 2.0: past Fable 5’s best result at a third of the cost The customer numbers hit harder than the benchmarks. Harvey matched Opus 4.8’s max-reasoning quality using 26% fewer tokens . Zapier’s churn workflow went from failing every prior model to 100% . A trading firm hit its best Opus score on a seventh of the reasoning tokens . Here is the free insight: this is a price-per-outcome launch wearing a model launch costume. Anthropic held the sticker price flat and moved everything underneath it. Intelligence up, tokens down, safety blocks down 85%, speed up 2.5x in Fast mode. Your bill for the same work drops even though the price page stayed identical. That only happens if you drive it correctly. Most people will run it on defaults and leave the upside sitting there. Behind the paywall, The Opus 5 Playbook : ▫️ what each setting costs and returns, and why running max is the standard way to overpayThe effort dial,▫️ what goes to Opus 5, what stays on Fable 5, what drops to SonnetThe routing table,▫️ worked examples on an $8K monthly billThe cost math,▫️ built around the one behavior every early tester namedSix copy-paste prompts▫️ agent builders should care about more than any benchmarkThe two beta features▫️ what opened up, what stays blocked, how the fallback worksThe safeguards map,▫️ by source model, with your loops intactThe migration checklist,▫️ including what still belongs on Fable 5The eval protocol and the honest limits, One subscription unlocks every system This is one build in a growing library. Premium opens all of them: ▫️ The Claude and Anthropic library https://www.the-ai-corner.com/t/claude-and-anthropic?r=1krivi ▫️ The AI Tools and Models library https://www.the-ai-corner.com/t/ai-tools-and-models?r=1krivi Plus 3 fresh systems every week. One correctly set effort dial pays the subscription back inside a week. 🧠 The Opus 5 Playbook The effort dial, the routing table, the cost math, six prompts, the migration checklist, and the honest limits, in one system. Get The Opus 5 Playbook below 👇 Try premium free for 7 days. Or get 50% off this week only. Keep reading with a 7-day free trial Subscribe to The AI Corner to keep reading this post and get 7 days of free access to the full post archives.