Anthropic launches new AI model with 'more honest' responses
What's the story
Anthropic has launched the latest version of its flagship AI model, Claude Opus 4.8. The new iteration comes with improvements in coding, reasoning, and general knowledge capabilities. However, the company claims that the biggest improvement lies in its honesty. Opus 4.8 is more likely to admit uncertainty and less likely to make unsupported claims than its predecessor models.
Enhanced performance
New model offers modest but tangible improvements
Anthropic has described Opus 4.8 as a "modest but tangible improvement" over its predecessor, Opus 4.7. The company claims that the new model performs better in agentic tasks and is sharper in judgment and more reliable during such tasks, according to early testers. This update comes just over a month after the release of the previous AI model by Anthropic.
User control
Opus 4.8 introduces Effort Control feature
Along with improved honesty, Opus 4.8 also brings a new feature called Effort Control. This lets users decide how much "thinking power" they want the AI to use before answering a question. If users opt for a high-effort setting, Claude will take longer to reason deeply before responding. A lower setting would make the AI answer faster while consuming fewer resources and rate limits.
Advanced feature
Dynamic Workflows also added in latest model
Another major addition in Opus 4.8 is Dynamic Workflows, which is currently available in research preview. The feature is aimed at handling extremely large and complicated tasks that can't be realistically handled by a single AI agent. It allows larger AI models like Opus to coordinate hundreds of parallel AI subagents at once, making it useful for tasks such as checking thousands of lines of code or hunting software bugs across an entire service.
Upcoming features
Mythos-class capabilities coming to customers soon, says Anthropic
Anthropic has also revealed that it has made rapid progress on safeguards and expects to bring Mythos-class capabilities to customers in the coming weeks. This comes after the company alarmed cybersecurity experts by revealing that Mythos AI was dramatically better at discovering vulnerabilities than previous models.