Unverified50% confidenceBenchmarkExact time
Among 100 additional test questions, Gemini 4 answered 50 correctly, got 8 wrong, and answered "don't know" to 42; GPT-6 Astra answered 63 correctly and got 19 wrong.
1
Sources
50%
Confidence
Short-term (~14 days)
Relevance
10/14/2026
First Seen
Valid until: 10/28/2026
Sources
Related Entities
Related Claims
UnverifiedGemini 3.7 Flash answered all 169 returned items correctly in the sufficiency task, but 23 calls yielded no result after three attempts; counting those as failures reduces the number of passing families from a perfect score to 28 out of 48.72% similarUnverifiedOn the Artificial Analysis index, Claude Opus 5.5 scores 58, while Gemini 4 Argon and GPT-6 Astra both score 53.71% similarUnverifiedAfter testing, programming blogger Theo pointed out that more than half of Gemini 4's judgments about his codebase were wrong, making it the only one of the 7 tested models with more than half errors.69% similarUnverified46个模型中有41个完整跑完全部28道题，3个因服务错误未完成，GPT-5.6 Terra因配额限制止步24/28，GPT-6 Astra只跑了1道题69% similarUnverifiedGPT-5.5 ran out of usage quota, leaving 35 of the 90 questions unfinished; all 55 answered questions were correct, but Kaggle counted the unanswered questions as wrong, causing its official score to be lower.67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1030260API
curl https://kongchang.com/api/v1/knowledge/claims/1030260MCP
get_claim(id=1030260)