Anthropic Drops Opus 4.8 With a Focus on Honesty
Anthropic's latest flagship model promises to admit what it doesn't know instead of faking it.
Anthropic just shipped Opus 4.8, the newest iteration of its most powerful publicly available AI model. The pitch? It's less likely to make stuff up.
The company says Opus 4.8 is better at flagging uncertainties in its own outputs and less prone to making unsupported claims. In other words, Anthropic is leaning hard into reliability over raw capability — a notable move in an industry still battling hallucination problems.
Pricing stays flat. Opus 4.8 costs the same as the previous 4.7 version, meaning users get the upgrade without reaching deeper into their wallets.
The release signals Anthropic's continued strategy of iterating quickly on its Claude model family while positioning itself as the "responsible" alternative in the frontier AI race. Whether the honesty improvements hold up under real-world pressure remains to be seen.