Claude Sonnet 5 API 是一款具备智能体能力的多模态接口,不仅提供接近 Opus 模型的性能,还支持原生计算机使用与自我验证功能。它引入了精细化的“思考强度 (Effort Levels)”,帮助工程团队在 API 部署成本与任务推理复杂度之间实现精准平衡。

Pricing: Input 170 credits / 1M tokens (≈ $0.850), Output 855 credits / 1M tokens (≈ $4.275) . High-tier top-ups include +10% bonus credits. Effective pricing is approximately 10% lower than the above rates.
claude-sonnet-5

Automatically append previous messages to maintain multi-turn context. May increase token usage.

Function calling
Thinking process
Max tokens
Stream Response

How can I help you today?

README

Complete guide to using claude-sonnet-5

专为自主智能体与复杂代码生成打造的 Claude Sonnet 5 API

Original image
自主智能体智能与多步骤任务执行
媲美 Opus 的性能与基于思考强度的成本控制
进阶代码生成与原生电脑操控
增强的企业级安全与稳健的风险管控
体验新一代内容创作工具。
智能体搜索性能与性价比曲线优化
分类Claude Sonnet 5Claude Sonnet 4.6Claude Opus 4.8
PositioningStrongest Agentic Mid-tier ModelPrevious Generation SonnetCurrent Flagship Opus Model
Context Window1M tokens1M tokens1M tokens
Agentic Capabilities★★★★★ (Best)★★★★★★★★★
Agentic Coding (SWE-bench Pro)63.20%58.10%69.20%
Agentic Coding (Terminal-Bench)80.40%67.00%82.70%
Computer Use (OSWorld-Verified)81.20%78.50%83.40%
Multidisciplinary Reasoning (Humanity's Last Exam)43.2% (no tools) / 57.4% (with tools)34.6% / 46.8%49.8% / 57.9%
Knowledge Work (GDPval-AA)161813951615
Adaptive ThinkingEnabled by DefaultSupportedSupported
SafetySignificantly improved, default cyber safeguardsGoodHighest
Cost-Performance★★★★★ (Best)★★★★★★★
Recommended Use CasesAgent Building, Large-scale Coding, Enterprise AutomationGeneral TasksExtremely Complex Deep Reasoning

Claude Sonnet 5 API 常见问题解答