Bottom line: Anthropic releases Sonnet 5, which approaches Opus 4.8 performance but is available at significantly lower prices, making autonomous agent tasks more cost-effective.
Anthropic has released Claude Sonnet 5, which significantly expanded the autonomous capabilities of its Sonnet class and closed the performance gap to Opus 4.8. For CTOs, this means: tool usage, code generation, and autonomous task planning are now available in the affordable model class.
Claude Sonnet 5 closes the performance gap between Sonnet and Opus. The new Sonnet model can plan independently, deploy tools such as browsers and terminals, and work autonomously – capabilities that were reserved for larger, more expensive models only a few months ago. Compared to its predecessor Sonnet 4.6, Sonnet 5 shows measurable improvements in critical agent tasks: reasoning, tool usage, code generation, and knowledge work.
Performance characteristics: Sonnet 5 achieves performance scores close to Opus 4.8, while Sonnet 4.6 lags significantly. At the same time, security assessments show that Sonnet 5 has a lower rate of undesired behaviors than its predecessor and is therefore safer to deploy in agent contexts. One limitation: Sonnet 5 performs significantly weaker on cybersecurity tasks than current Opus models – an important difference that CTOs should consider when choosing a model.
Pricing: Sonnet 5 costs 3 US dollars per million input tokens and 15 US dollars per million output tokens in the regular phase (starting September 1, 2026). During the introductory phase until August 31, 2026, the price is 2 and 10 US dollars respectively. Opus 4.8 costs 5 and 25 US dollars by comparison. The model family thus shows a coherent performance range: Sonnet 5 offers high capabilities at low cost, Opus 4.8 higher accuracy at higher prices.
Availability: Sonnet 5 is now available through all Claude plans – as standard for Free and Pro, and available for Max, Team, and Enterprise. Developers access it via the Claude API using the model identifier `claude-sonnet-5`. Model selection now follows a clear pattern: total cost of ownership across the effort level, not absolute performance ceilings.
Source: www.anthropic.com · Published June 29, 2026
Lumi AI News — AI-assisted curation in accordance with Art. 50 EU AI Act. Paraphrase and classification through Lumi News Pipeline v1.7.2.