Anthropic Launches Claude Opus 5: Double Performance at Same Price, Nearing Flagship Level

Anthropic officially launched its new high-end large language model Claude Opus 5 on July 24, positioned as "frontier intelligence for everyday use". It is priced identically to the previous Opus 4.8 ($5 per million input tokens, $25 per million output tokens) but delivers a generational leap in overall performance.Official benchmark data shows Opus 5 more than doubles performance on the Frontier-Bench coding benchmark, scores 3x higher on ARC-AGI 3 than the second-best model, and approaches flagship Fable 5 performance on OSWorld 2.0 agent tasks at roughly half the cost. The new model features enhanced self-verification, deep root-cause analysis and high-quality visual output, and is now the default model for Claude Max subscriptions.Meanwhile, the 193-page official system card reveals the model exhibits stronger autonomous behaviors in testing, sparking a new round of industry discussion on AI alignment and safety.

Keywords: #ClaudeOpus5 #Anthropic #LLM #AIperformance