Anthropic has released Claude Sonnet 5.5, its latest AI model in the Claude 5.5 family. The new model generates output over 30 percent faster and costs up to 30 percent less per task compared to previous iterations, while nearly matching the flagship Opus 5.5 across multiple benchmarks. Designed for everyday technical tasks like documentation and coding, Sonnet 5.5 maintains identical per-token pricing to Sonnet 5 at $2 per million input tokens and $10 per million output tokens, achieving cost savings by using fewer total tokens per task.
According to Anthropic, the model’s most significant gains appear in software engineering. On Terminal-Bench 4.0, an agentic coding benchmark, Sonnet 5.5 reached 70.6 percent compared to Sonnet 5’s 10.3 percent. On the CursorBench 4.0 coding test, it scored 55.5 percent, landing just behind Opus 5.5’s 57.8 percent score. On the multi-industry knowledge-work benchmark GDPval-AA, Sonnet 5.5 scored 1,844 points, placing it virtually on par with Opus 5.5 (1,846) and above OpenAI’s GPT-6 Sol (1,487).
Alongside the model release, Anthropic announced plans to launch Claude Haiku 5.5 in the coming weeks to target high-throughput, low-cost use cases. Sonnet 5.5 also introduces enhanced vision and tool-use capabilities, including increased batching of tool calls and UI generation capabilities.
Why it matters
Sonnet 5.5 matches top-tier performance on coding and knowledge benchmarks at lower token usage, offering founders cost-effective frontier capabilities.
Anthropic’s frequent tool-call batching reduces total required execution steps, lowering latency and operational costs for agentic developer workflows.
Tiered pricing competition between Anthropic’s Claude 5.5 suite and OpenAI’s GPT-6 models provides enterprise operators leverage to optimize AI infrastructure spend.
Source: the-decoder.com



