Anthropic isn’t ready to let regular users look at its supposedly super-powerful Claude Mythos AI model just yet. But the AI company has just released an upgrade to its flagship product, Claude Opus — now in its 4.8 version.
“It builds on Opus 4.7 with improvements across benchmarks, and is a more effective collaborator,” Anthropic promised in a press release Thursday. Indeed, the benchmark numbers, below, show very minor improvements across the board.
One major improvement, allegedly, is in the area of hallucinations. Claude Opus 4.8 won’t lie to users as much. “Early testers report that Opus 4.8 is more likely to flag uncertainties about its work and less likely to make unsupported claims,” Anthropic said, touting the model’s “honesty.”
Claude Opus 4.8 has ‘better judgment’
“Claude Opus 4.8 has noticeably better judgment,” an engineer at Shopify, Tom Pritchard, told Anthropic. The coding version of the model “asks the right questions, catches its own mistakes, and pushes back when a plan isn’t sound.”
Support authors and subscribe to content
This is premium stuff. Subscribe to read the entire article.
