Unverified50% confidenceOpinionExact time
After 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.
1
Sources
50%
Confidence
Short-term (~14 days)
Relevance
10/14/2026
First Seen
Valid until: 10/28/2026
Sources
Related Entities
Related Claims
Unverified在285次实测运行中，Gemini取得283次语义通过、2次语义部分通过、0次可复现语义失败69% similarUnverifiedAmong 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.69% similarUnverifiedIn a test of 360 clean texts, Jev made only 1 misjudgment, gpt-5-mini made 1 misjudgment, Gemini 2.5 Flash-Lite and Gemma each made 2 misjudgments, gpt-oss-safeguard-20b made 4 misjudgments, and Llama Guard 4 made 26 misjudgments.68% similarUnverifiedAll Claude, GPT-6.1, and Gemini models achieved a perfect score of 12/12 in the experiment, with only 1 accidental impersonation response out of 252 attempts.66% similarUnverifiedGemini 4 Argon reached 51.3% in the workflow automation test, while the second place had only 42.5%.66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1030258API
curl https://kongchang.com/api/v1/knowledge/claims/1030258MCP
get_claim(id=1030258)