Claude Opus 4.8

The highest-tier Opus model at release. Emphasizes reliability and coding honesty — reported to be approximately four times less likely than Opus 4.7 to let flaws in self-authored code pass unremarked. Highest score on Anthropic's Legal Agent Benchmark at launch. First model to complete every case on Anthropic's Super-Agent benchmark. Supports adaptive thinking (auto-adjusts reasoning depth). Fast mode available at 2x standard pricing. Priced at $5/$25 per million input/output tokens. Succeeded as the top flagship by Claude Fable 5 on June 9, 2026.