Anthropic Launches Claude Opus 5, Its New Flagship Model

09:00 · July 29, 2026

Claude Opus 5 is now Anthropic’s default model on Claude Max and its strongest available model on Claude Pro, arriving with a 1 million token context window and a 128,000 token maximum output. Anthropic held pricing flat at $5 per million input tokens and $25 per million output tokens, the same rate as Opus 4.8, while reporting a generational jump in capability. On Frontier-Bench v0.1, the company’s internal benchmark for coding and knowledge work, Opus 5 more than doubles Opus 4.8’s score and comes out ahead of every other model Anthropic tested it against. On SWE-bench Verified, the industry’s standard software engineering benchmark, Opus 5 scored 96.0 percent, which Anthropic describes as close to saturating the test. The model also leads on ARC-AGI 3 and on OSWorld 2.0, a computer-use benchmark, where it beats rival flagship models at roughly a third of their cost per task. Beyond raw capability, Anthropic is emphasizing safety: internal evaluations show Opus 5 has the lowest rate of deceptive or manipulative behavior of any Claude model released to date, a distinction Anthropic argues matters more as these models are increasingly deployed as autonomous, long-running agents handling real business workflows rather than answering one-off questions.