Anthropic: Claude Opus 4

Claude Opus 4 is Anthropic's closed-weight model accepting text, images, and files as input, with a 200,000-token context window and up to 32,000 tokens of output per response. It supports tool use and reasoning, though structured output support is unconfirmed from available data. The large context window makes it viable for tasks involving long documents or extended multi-turn sessions. At $15.00 per million input tokens and $75.00 per million output tokens, this is an expensive option that needs to justify its cost against benchmarks. Its blended score of 70.8 covers only 3 benchmarks, including a 72.0 on coding and a notably lower 51.1 on the AI Index, so its performance profile is not yet broad enough to draw firm conclusions. Buyers running high-volume workloads will feel the output cost acutely, making this most sensible for lower-frequency, context-heavy tasks where the 200k window is a genuine requirement.