Anthropic Ships Claude Opus 5—Near-Fable Intelligence at Half the Price
Claude Opus 5 arrives at Opus 4.8 pricing with claimed state-of-the-art coding scores, near-Fable 5 intelligence at half the cost, and record alignment marks.
In Brief
- Claude Opus 5 launches at $5 per million input tokens and $25 per million output—the same price as Opus 4.8, which Anthropic says buys near-Fable 5 intelligence at half the cost
- Anthropic claims new state-of-the-art results on Frontier-Bench and GDPval-AA, with an ARC-AGI 3 score three times the next-best model, though Opus 5 stays behind Mythos 5 on cybersecurity
- Anthropic’s automated behavioral audit rates Opus 5 its most aligned model to date at 2.3 on misaligned behavior, with cyber classifiers expected to intervene about 85% less often than on Fable 5
Anthropic released Claude Opus 5 on Thursday, pitching it as “a thoughtful and proactive model that comes close to the frontier intelligence of Claude Fable 5 at half the price,” according to the company’s announcement. Pricing holds at $5 per million input tokens and $25 per million output tokens—identical to its predecessor, Opus 4.8.
Anthropic says Opus 5 sets a new state of the art on coding and knowledge-work evaluations including Frontier-Bench and GDPval-AA, more than doubling Opus 4.8’s Frontier-Bench v0.1 performance at a lower cost per task. Those Frontier-Bench v0.1 results come from an internal run on the mini-SWE-agent harness and a GKE backend, with mean reward over 5 attempts per task. Opus 4.8 served as fallback on safety-classifier refusals for Opus 5 and Fable 5.
The claims drew immediate ecosystem uptake: GitHub made Opus 5 available in Copilot the same day, describing it as designed for complex, long-running coding tasks. Anthropic is also making Opus 5 the default model on Claude Max and the strongest model available on Claude Pro, while one caveat it volunteers itself is that Opus 5 “remains behind Mythos 5 on cybersecurity tasks.”
What Claude Opus 5 delivers for the money
The benchmark spread is broad. On ARC-AGI 3, a novel problem-solving evaluation, Anthropic says Opus 5’s score is three times as high as the next-best model. On Zapier AutomationBench it passes around 1.5 times as many end-to-end business tasks as the next-best model at the same cost, and on the OSWorld 2.0 computer-use benchmark it surpasses Fable 5’s best result at just over a third of the cost. On CursorBench 3.2 at max effort, the model lands within 0.5% of Fable 5’s peak score at half the cost per task.
Early-access partners back the framing in testimonials published with the announcement. “Claude Opus 5 delivers near Fable 5 intelligence at Opus speed and cost,” Cursor’s team said. Zapier reported the model “topped Zapier’s AutomationBench leaderboard without spending more tokens than prior Claude models,” running a full churn-prevention workflow end to end: “Previous models didn’t pass; Opus 5 hit 100%.” Box found Opus 5 beats Opus 4.8 by 8% overall, with an 11% gain in data analysis and 17% in due diligence workflows.
Science is the quieter gain: Anthropic reports improvements on every internal life-sciences evaluation, including a 10.2-percentage-point jump on inferring molecular structures from spectroscopy data and 7.7 points on predicting protein-variant function. The launch extends the cadence Anthropic set this month after capping Fable 5 usage and upgrading Claude’s voice mode a day earlier.
Alignment numbers, cyber guardrails and the fine print
Anthropic’s automated behavioral audit scores Opus 5 at 2.3 on overall misaligned behavior—the lowest of its recent models—making it, by the company’s own measure, its most aligned model to date, with the lowest rates of deceptive behavior and the least susceptibility to being tricked into misuse.
The safety architecture is deliberate: Anthropic says it intentionally avoided training Opus 5 on cyber tasks, and while the model nears Mythos 5 at finding vulnerabilities on OSS-Fuzz, it stays far behind at developing exploits. Cyber classifiers block binary-based vulnerability scanning, penetration testing, and exploit generation, but Anthropic expects them to intervene around 85% less often than on Fable 5, with flagged requests falling back to Opus 4.8 by default. Details live in the company’s System Card.
Alongside Opus 5, Anthropic is shipping two updates in beta. On the Claude Platform, developers can now change which tools Claude can use mid-conversation without invalidating the prompt cache. On the API, users can now choose to have requests flagged by the safety classifiers on Opus 5 (or Fable 5) automatically route to another model; with automatic fallbacks on, API requests always route to the best available model by default rather than being blocked. A Fast mode runs at roughly 2.5 times default speed for twice the base price, available on the Claude Platform and through usage credits in Claude Code. Consistent with prior Opus models, Opus 5 does not have data retention requirements for general access. For guidance on getting the most out of the model, Anthropic published a prompting guide for Opus 5. The competitive stakes are high enough that Washington is watching the model family itself, as Frontierbeat covered when the White House accused Moonshot of distilling Anthropic’s Fable.
FAQ
How much does Claude Opus 5 cost?
$5 per million input tokens and $25 per million output tokens—unchanged from Opus 4.8—with a Fast mode at twice the base price running about 2.5 times the default speed.
How does Opus 5 compare with Fable 5 and Mythos 5?
Anthropic says it comes close to Fable 5’s frontier intelligence at half the price and beats it on cost-per-task across several benchmarks, but remains behind Mythos 5 on cybersecurity and biology research tasks.
What safety changes ship with Opus 5?
It is Anthropic’s most aligned model per its behavioral audit (2.3 misalignment score), cyber classifiers intervene about 85% less often than on Fable 5, and flagged requests fall back to Opus 4.8 or another model via optional API auto-fallbacks.
