Unverified50% confidenceBenchmarkExact time
在 skate bench 测试中，Max 模式相比 X high 准确率仅从 78% 提升到 79%，但成本高了 13 倍
1
Sources
50%
Confidence
Short-term (~14 days)
Relevance
9/25/2026
First Seen
Valid until: 10/9/2026(expired)
Sources
Related Entities
Related Claims
Unverified在 Theo 的 skate bench 测试中，X high 档平均每次响应 338 个 token、耗时 6 秒、最慢 31 秒76% similarUnverified通义千问3.6在SWE-Bench测试中得分78.8分，对比Claude Opus的80.9分63% similarUnverifiedDevin在SWE-bench评测中达到13.86%的解决率（当时最高）63% similarUnverified在Terminal Bench 4.0最高精度测试中，Fable 5.1准确率为55.8%，成本为19.50美元；Astra准确率为56.7%，成本为10.35美元62% similarUnverified在CursorBench 4.0基准中，Grok 4.7 Extra High得分46.3%，逼近Opus 5 Max的46.6%，成本仅6.01美元62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/948905API
curl https://kongchang.com/api/v1/knowledge/claims/948905MCP
get_claim(id=948905)