Claude Sonnet 5 Review

Anthropic's mid-tier model, released on 30 June 2026 between Fable 5 and Opus 5 in the same generation. It carries the same million-token context window as the models above it and the same 128K output ceiling, at $2 per million input tokens and $10 per million output, which is roughly forty percent of the Opus rate and a fifth of Fable's. The API surface matches the rest of the current family: adaptive thinking with effort levels from low through max, no fixed token budget, no assistant prefill, no sampling parameters.

One notable omission is mid-conversation system messages, which the Opus and Fable models accept and Sonnet 5 rejects outright, so operator instructions injected partway through a conversation need a different approach here. It is the obvious default for high-volume production traffic where the frontier tier is not worth its multiple. This grade is a provisional placeholder pending hands-on testing.