Anthropic today announced Claude Sonnet 5.5, the second model in its Claude 5.5 family, positioning it as a faster, cheaper workhorse for coding, scoped tasks, and routine business work. The model arrives less than a week after the company released flagship Opus 5.5.
According to the announcement, Sonnet 5.5 represents a clear upgrade over Sonnet 5, running more than 30% faster and costing up to 30% less for most work. The pricing and speed improvements target high-volume everyday AI usage, from slides and spreadsheets to code generation.
Anthropic is splitting its model lineup into distinct tiers: frontier intelligence with Opus 5.5 and practical utility with Sonnet 5.5. By making the workhorse model significantly faster and cheaper, the company signals that competition in AI is shifting from raw benchmark scores to cost efficiency and real-world throughput. Most developers and businesses don't need the most capable model — they need one their finance team will approve. Sonnet 5.5 aims to fill that role, putting pressure on rivals to match on price rather than just capability.
The launch also follows CEO Dario Amodei's recent call to "pace the frontier" — an argument that the industry should slow capabilities development for safety. Releasing a cheaper, faster model shows that slowdown is not a halt on selling AI; it's a strategy for capturing the mid-market while the frontier debate continues.
Artificial Analysis reports that Sonnet 5.5 scores 56 on its Intelligence Index at maximum effort, two points behind Opus 5.5. It reaches 64% on Terminal-Bench 4.0, compared with 60% for Opus 5.5 and GPT-6 Astra. The evaluations used a pre-release deployment with a structured-output bug that Anthropic says is fixed in the public release; Artificial Analysis plans to rerun relevant evaluations.
That performance comes with a usage caveat: Artificial Analysis measured roughly 193,000 output tokens per task at maximum effort and a $7.60 cost per task, around 50% higher than Sonnet 5. Its reported API pricing remains $2/$10 per million input/output tokens. Anthropic’s claim of lower cost for most work therefore should not be read as a guarantee of lower cost at every effort setting.
BridgeMind highlights Sonnet 5.5 beating Opus 5.5 on agentic coding; the benchmark comparison is specific to the measured task, rather than a claim of superiority across all coding work. Boris Cherny shares a Claude Code bug-fixing demo and describes it as 30% faster with 30% less usage.
Sources: Source 1 · Source 2 · Artificial Analysis · @bridgemindai · @bcherny