Anthropic's Opus 5 nearly matches its flagship Fable model at half the price

[Darren51]

Anthropic released Claude Opus 5, describing it as a thoughtful and proactive everyday model that comes close to the frontier intelligence of Fable 5 at half the price. On coding and knowledge work evaluations like Frontier-Bench and GDPval-AA, Opus 5 is the new state of the art among Anthropic's models, though it remains behind Mythos 5 specifically on cybersecurity tasks. It is now the default model on Claude Max and the strongest model available on Claude Pro

The release is Anthropic's fourth Claude 5 series model in under two months, following Opus 4.8, Sonnet 5, Mythos 5 and Fable 5, a pace Axios described as evidence that AI deployment has shifted from blockbuster launches to rapid, continuous improvement on capability, cost and speed. On ARC-AGI 3, a benchmark testing novel problem solving, Opus 5 scored three times as high as the next best model, and on the OSWorld 2.0 computer use benchmark it surpassed Fable 5's best result at just over a third of the cost

Anthropic says its pre deployment automated behavioral audit found Opus 5 to be its most aligned model to date, with the lowest rates of deceptive behavior and the least susceptibility to being tricked into misuse. Pricing stays the same as its predecessor Opus 4.8 at 5 dollars per million input tokens and 25 dollars per million output tokens, and the launch includes a new opt in Automatic Fallbacks feature that reroutes any request blocked by a safety classifier to another model instead of simply returning a refusal

Introducing Claude Opus 5www.anthropic.com

[Ellie85]

Three times the next best model on ARC-AGI 3 is a huge gap for a benchmark specifically designed to resist memorization

[Wrench]

Four Claude 5 series releases in under two months is a pace that would have been unthinkable for a single lab just a couple years ago

[NatureBoyOwen16]

The wind tunnel and cell artifact demos are a nice touch, visual output quality rarely gets its own section in a model announcement like this

[Teal Sparrow]

Coming in at a third of Fable 5's cost on OSWorld while beating its best score is the number that will actually move enterprise buying decisions

[WearyCoder]

Most aligned model to date claims always deserve some skepticism until independent researchers get to poke at it themselves

[Cole99]

The Automatic Fallbacks feature feels like a direct response to all the recent complaints about guardrails just returning blunt refusals

[Inlet]

Staying behind Mythos on cybersecurity specifically while beating everything else is a sensible segmentation of risk across the model lineup

[Josh_79]

Same price as Opus 4.8 but this much of a capability jump is a real value story for anyone already on that tier

[Memory Jamie]

The effort dial letting users trade intelligence for cost and speed is becoming a really useful pattern across the whole industry now