Unverified85% confidenceFactTime unknown
Claude Sonnet 4.5在软件工程基准测试中大幅超越了前代Sonnet 4模型，甚至超过了定位更高端的Opus模型
1
Sources
85%
Confidence
Medium-term (~90 days)
Relevance
6/1/2026
First Seen
Valid until: 8/30/2026
Sources
Claude 4.5 Sonnet实测：一条指令构建完整AI视觉应用
bilibiliUltralytics
Related Entities
Related Claims
Unverified在计算机使用和世界知识任务上，Sonnet 5 已接近 Opus 4.8 的水平，但在智能体编程领域仍明显落后于 Opus 4.877% similarUnverifiedClaude 4.0（Opus/Sonnet）在复杂推理和工具调用上相比前代实现了显著提升76% similarUnverifiedClaude Haiku 4.5在编程性能上已能与上一代旗舰模型Sonnet 4匹敌74% similarUnverified在前端编码任务上，Mimo V2.5 Pro已经与Claude Sonnet 4.6处于同一性能区间74% similarUnverifiedSonnet 4.8预计将继承Opus 4.7引入的多项重大升级，包括UI原型、截图和复杂架构图的识别准确率预计超过98%73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/24171API
curl https://kongchang.com/api/v1/knowledge/claims/24171MCP
get_claim(id=24171)