Unverified50% confidenceOpinionExact time
The blogger pointed out four discount factors in the benchmark comparison: different test frameworks (TerminalBench 1017 ran within the ClaudeCode framework while the baseline ran within the DeepSeek framework), Artificial Analysis's own test of the old Flash on TerminalBench scored only 26.8% rather than 31.2%, the competitor comparison used Opus 5 and GPT 5.6 Sol rather than the latest models, and the old model's NL2 repo score is inconsistent across different tables (64.0 vs 65.4).
1
Sources
50%
Confidence
Short-term (~14 days)
Relevance
10/18/2026
First Seen
Valid until: 11/1/2026
Sources
DeepSeek New Version Hands-On: Same Price, Costs Slashed by 60%
bilibili枫叶边城10/18/2026
Related Entities
Related Claims
Unverified盘古2.0 Flash在SWE-Bench Verified工程实战测试中得分仅63.1，在所有对比模型中垫底75% similarUnverifiedOn the Terminal Bench 2.1 benchmark, GPT-5.6 Sol's Ultra mode approaches 92%, surpassing the competing product's performance of 88%.67% similarUnverified盘古2.0 Flash在Tau2-Bench客服工具调用测试中得分88.0，在WildCloudBench中得分41.5超过397B模型的34.567% similarUnverified盘古2.0 Flash在可验证指令遵循测试中得分95.9，位列全场第一65% similarUnverifiedIn the new version of the 24-point problem (numbers 2, 7, 8, 10, requiring three different solutions), DeepSeek V4.1 Flash got all three solutions wrong, each violating the constraint that each number can only be used once.64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1059843API
curl https://kongchang.com/api/v1/knowledge/claims/1059843MCP
get_claim(id=1059843)