Unverified50% confidenceBenchmarkExact time
DeepSeek-V3发布时在多项基准测试中与Claude 3.5 Sonnet持平甚至超越，训练成本据报道仅为主流闭源模型的数分之一
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/7/2026
First Seen
Valid until: 10/5/2026
Sources
Anthropic如何一步步失去开发者信任：定价、透明度与供应商锁定
hackernewshackernews7/6/2026
Related Claims
Unverified经过对比测试，DeepSeek V4 Pro的编程效果甚至超过了Claude Sonnet，而且价格便宜得多75% similarUnverifiedGLM-4.7在代码能力基准测试中超过DeepSeek V3.2及Claude Sonnet 4.571% similarUnverified一些开源或轻量级模型（如Mistral、Llama系列通过第三方托管）的调用成本可能只有Claude 3.5 Sonnet的十分之一甚至更低70% similarUnverifiedDeepSeek与Claude Sonnet在SWE-bench编码基准测试上的差距仅在1%以内67% similarVerifiedClaude 3.5 Sonnet在SWE-bench等代码基准测试中长期领先66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/170418API
curl https://kongchang.com/api/v1/knowledge/claims/170418MCP
get_claim(id=170418)