-
$1.8 Billion Pledged for Biodiversity at COP17 as Monaco Reaffirms Commitment
-
Karen Wazen Launches First Standalone Boutiques in Cairo
-
Phoenix Symphony Resumes Performances Amid Financial Uncertainty
-
Municipal Elections: How Service Failure Becomes a Public Health Crisis in South Africa
-
Turkish Energy Minister Highlights Qatar as Key Partner for Regional and Global Energy Security
-
Rene Eyre Revisits Autobiographical Performance 'Say A Little Prayer' in London
-
London Museum Granted 24/7 License Ahead of November Opening
-
UAE's HCT Holds Record Graduation and Announces Major Training Expansion
-
Dana White Confirms Flight to Washington for Trump Meeting on Press Secretary Role
-
NASA Awards $10B Lunar Contract Framework to SpaceX, Blue Origin
-
Dengue Fever Expands Into High Himalayas as Climate Change Warms Nepal
-
Katy Perry Tells Fans Ex-Husband Russell Brand 'Should Be in Jail' During UK Concert
-
IoT Refrigerator Hacks Disrupt Blood Labs and Supermarkets Across Europe and US
Anthropic bets on cheaper AI with new model
Anthropic on Friday released Claude Opus 5, a new artificial intelligence model the company says approaches the performance of its most powerful system at half the price.
The San Francisco-based startup, a chief rival to OpenAI, said the model tops industry benchmarks for coding and knowledge work while consuming fewer computing resources than earlier versions.
The claim addressed the rising concerns of business customers who have balked at the expense of running advanced AI systems.
Claude competes with OpenAI's ChatGPT and Google's Gemini, and customers pay based on the amount of work performed, measured in units called tokens.
Opus 5 costs $5 per million input tokens and $25 per million output tokens, half the price of Fable 5, Anthropic's most expensive publicly available offering.
The company said the new model performs close to Fable 5 across many tasks despite the lower cost.
Cost has becomes a major issue after Anthropic and its archrival OpenAI, the maker of ChatGPT, both filed for stock market listings in June.
Wall Street is scrutinizing their business performance more closely than ever ahead of their market debuts.
On cybersecurity, Anthropic said it deliberately avoided training Opus 5 on cyber tasks, and that the model remains behind Mythos 5 -- its most capable and most tightly restricted system, available only to select partners -- on both offensive cyber work and biological research capabilities that could be misused.
Mythos is Anthropic's most advanced series of AI models, and their strength at finding weaknesses in computer systems -- and at turning those weaknesses into working attacks -- has caused controversy and pushed the company to build tight guardrails around them.
The company said Opus 5 finds software flaws about as well as Mythos 5 but lags far behind in turning those flaws into something that could be weaponized into an attack.
Citing the reduced risk, Anthropic said its automated safety filters will intervene roughly 85 percent less often on Opus 5 than on Fable 5, the Mythos-class model it released last month.
Fable 5 and Mythos 5 were briefly pulled from public access in June after the US Commerce Department imposed export controls, which were lifted at the end of that month.
A.Motta--PC