Anthropic announced its new AI model, "Claude Sonnet 5.5," on September 28. It claims to have increased output speed by over 30% and reduced the cost for the same tasks by up to 30% compared to the previous Sonnet 5. The API pricing remains unchanged at $2 per million input tokens and $10 for output.
In evaluations by AI benchmarking institution Artificial Analysis, Sonnet 5.5, when set to its highest "Max effort," scored 56 points on the "Intelligence Index." This marks an 18-point increase from Sonnet 5, bringing it within 2 points of the top-tier model, Opus 5.5, which scored 58 points.
Compared to OpenAI's GPT-6 Astra, Sonnet 5.5 surpasses it in agent performance. In Artificial Analysis's "Terminal-Bench 4.0," Sonnet 5.5 achieved 64%, while Astra scored 60%. For practical capability benchmarks like "GDPval-AA" and "AA-Briefcase," Sonnet 5.5 reached nearly the same level as Opus 5.5.
However, Sonnet 5.5 uses an average of approximately 193,000 output tokens per task at "Max effort." This is the highest measured by Artificial Analysis, about seven times more than Astra's Max setting. The cost per task is $7.60.
Anthropic has launched Claude Sonnet 5.5: it scores 56 on the Artificial Analysis Intelligence Index, just 2 points behind Opus 5.5 (max), but at the highest Output Tokens per Task we’ve seen With max effort, Sonnet 5.5 gains 18 points over Sonnet 5 and to #2 on the Intelligence… pic.twitter.com/aunF6RTycN — Artificial Analysis (@ArtificialAnlys) September 28, 2026
