Anthropic Launches Sonnet 5.5 With 30% Speed Boost and Lower AI Costs
Anthropic launches Sonnet 5.5, claiming 30% faster performance, lower token use, stronger agentic coding and cyber capabilities comparable to Anthropic's Opus 5.
Summary
Anthropic released Sonnet 5.5 on September 28, 2026, positioning its mid-tier model as a faster, significantly cheaper assistant for everyday coding and office documents. It claims a 30% speed gain over Sonnet 5 and significantly slower token consumption. Sonnet 5 launched about three months earlier around efficient, lower-cost agent deployment. Although Sonnet is less powerful than Opus, Anthropic says its agility suits some workloads. Its benchmarks place Sonnet 5.5 above Opus 5.5 in agentic coding, likely because it can spawn multiple agents without exceeding cost limits.
Anthropic describes Sonnet 5.5's cyber capabilities as comparable to Opus 5, making it the first Sonnet subject to the same cyber safeguards as Fable and Opus. A new Haiku, Anthropic's smallest model, is planned within weeks, but no firm date was given. In the previous week, OpenAI introduced several models, including enhanced versions of mid-tier Sol and budget-focused Luna, while Meta announced a model intended to power an upcoming smart-glasses feature.
Positives
- 30% faster performance than Sonnet 5 could shorten everyday coding and office-document workflows.
- Significantly slower token consumption promises lower operating costs than Sonnet 5.
- Sonnet 5.5 beat Opus 5.5 on Anthropic's agentic-coding benchmarks while remaining within multi-agent cost limits.
- Cyber capabilities comparable to Opus 5 extend higher-tier functionality to Anthropic's mid-tier model.
- A new version of Haiku, Anthropic's smallest model, is planned for release within weeks.
Risks & concerns
- Sonnet 5.5 remains less powerful overall than Anthropic's Opus tier.
- Opus 5-level cyber capabilities require Sonnet's first application of safeguards previously used for Fable and Opus.
- Anthropic provided no firm release date for the coming Haiku update.
- The 30% speed gain and benchmark advantage are Anthropic's own performance claims.
- OpenAI's enhanced Sol and Luna models and Meta's smart-glasses model intensify competitive pressure.