Claude Opus 4.8: "a modest but tangible improvement"
Claude Opus 4.8 reduces hallucinations by being four times less likely to let code flaws pass unremarked
“Users will find Opus 4.8 to be a modest but tangible improvement on its predecessor. There's still more to be done: we're working on developing and releasing models that provide many of the same capabilities as Opus at a lower cost.”
Anthropic released Claude Opus 4.8 with a focus on honesty and reduced hallucinations — the model is four times less likely than its predecessor to let code flaws pass unremarked, and achieves lower incorrect rates by abstaining rather than guessing. Pricing is unchanged at $5/$25 per million tokens with the same 1M context window. A notable new engineering feature is mid-conversation system messages, enabling cheaper agentic loops by preserving prompt cache hits on earlier turns.