Claude Sonnet 5.5: the workhorse model that finally earns the middle seat

Anthropic’s Claude Sonnet 5.5, released late Monday, is priced like Sonnet 5 (US$2 / US$10 per million input/output tokens) but claims to finish most tasks with fewer tokens, run more than 30 per cent faster, and land far closer to Opus 5.5 on coding and knowledge-work benchmarks. On Terminal-Bench 4.0 it jumps to 70.6 per cent from Sonnet 5’s 10.3 per cent; early design and document testers praise clearer writing and less hand-holding. Slack, Zendesk, Epic and Unity are among firms quoting faster iteration and fewer wasted tool calls in early trials.

In practice that positions Sonnet 5.5 as the default for bug fixes, slide decks, spreadsheets and well-scoped agent loops, while Opus stays the pick for open-ended judgment calls. Cyber capabilities are high enough that Anthropic shipped Opus-style cyber fallbacks on a Sonnet for the first time — routine coding is fine, higher-risk security prompts may drop to Sonnet 5. Migration notes matter: thinking-disabled and forced-tool modes need updates before teams flip production traffic.

Fair take: if your stack already lives on Claude, Sonnet 5.5 is an easy upgrade for cost-sensitive volume work. If you need maximum judgment on ambiguous briefs, keep Opus in the loop. Canadian ecommerce and SaaS teams running support bots or codegen agents should A/B on their own traces — vendor benches are strong, but token bills and tool-call discipline are what hit the P&L.