Claude Opus 4.8

Opus 4.8 handles the work people actually bring to a workspace: multi-step reasoning, agentic coding across a whole repository, and synthesizing long, messy context into something coherent. New workspaces started on it until Opus 5.5 took its place, and a workspace already on it stays on it until someone changes its model.

Newer Anthropic models now sit above it. Opus 5 costs exactly the same and is better at all of the above, and Opus 5.5 costs a fifth less than either. Fable 5 and Fable 5.1 are stronger still, at double the price of Opus 5.

Best for: hard reasoning, agentic coding, long-document synthesis, anything where a wrong answer is expensive.