Anthropic Launches Claude Sonnet 5.5 With Matching Mid-Tier Pricing

Anthropic introduced Claude Sonnet 5.5 on Monday. Pricing holds steady at $2 per million input tokens and $10 per million output tokens, identical to the prior Sonnet 5 release and half the rate charged for Opus 5.5.

Benchmark results place Sonnet 5.5 ahead of Opus 5.5 on Terminal-Bench 4.0. Company data showed a 70.6 percent completion rate for the new model versus 66.4 percent for Opus 5.5. Separate evaluations from Artificial Analysis produced 63.6 percent for Sonnet 5.5 against 59.6 percent for Opus 5.5.

Artificial Analysis testing recorded a 64 percent mark on Terminal-Bench 4.0, reflecting a 50-point rise from Sonnet 5. On GDPval-AA, which assesses professional tasks across 44 fields via Elo ratings, Sonnet 5.5 reached 1844 while Opus 5.5 reached 1846.

Speed gains exceed 30 percent relative to Sonnet 5. Anthropic characterized the model as strongest on routine scoped work, bug repair, and generation of refined documents, slides, and spreadsheets, while also highlighting strong design judgment.

High-effort operation drives elevated token counts. Artificial Analysis logged roughly 193,000 tokens per task at maximum effort, the largest volume observed and about 60 percent above Opus 5.5, yielding an estimated $7.60 per task or roughly 50 percent more than Sonnet 5.

Medium-effort defaults preserve the projected cost reductions. The company stated that Sonnet 5.5 surpasses Sonnet 5's top coding result at under one-tenth the expense when run at the default Medium setting. Opus 5.5 continues to lead on complex tasks that require extended judgment. Claude Haiku 5.5 arrives in the coming weeks for high-volume uses.