human announced Friday its latest artificial intelligence model, Claude Opus 5. The company says it is the best performing and most cost effective product across many industry benchmarks.
According to the blog post, the Opus 5 outperformed the last model Anthropic released, the Claude Fable 5, in coding and knowledge work ratings, and the company said it was “designed for everyday use.” Opus 5 is also half the price of Fable 5, costing users $5 per million input tokens and $25 per million output tokens.
Anthropic said the Opus 5 is not state-of-the-art for “dangerous dual-use functions” like cybersecurity.
The model’s launch comes as Anthropic and its main rival, OpenAI, work to appease a more cost-conscious customer base where companies are reluctant to try expensive models without a clear picture of the return on investment. Anthropic also faces pressure from competitors promoting cheaper products. microsoft, Amazon, googleand many Chinese startups.
“Companies are looking for value in our feedback and in our customer base,” Diane Penn, head of research product management at Anthropic, said in an interview with CNBC. “Even if it’s a cheaper model or a cheaper product, if it’s not achieving the same level of quality, it doesn’t really help.”
Anthropic has been under intense scrutiny since April, when it released a model called Claude Mythos Preview and made it available to a limited number of users. The company said the model is very good at cybersecurity and is good at finding and exploiting vulnerabilities within software.
In June, Anthropic released an updated version of the model, Mythos 5, and also released another version of the model, Fable 5, which it says is safe for general use. According to the company’s blog post, the Fable 5’s features exceed those of all other models released to date.
Anthropic had to temporarily disable access to Mythos 5 and Fable five days after they were released to comply with an export control directive from the government citing “national security authorities.” After nearly two weeks of intense negotiations, export restrictions were lifted and Anthropic was able to regain access to both models.
Penn said Anthropic continues to work with government agencies, including on Opus 5, as part of pre-deployment and testing. The company said Opus 5 still lags behind Mythos 5 in offensive cybersecurity and biology research.
WATCH: Human partnership could boost Google’s revenue, says DA Davidson’s Gil Luria
