Unverified50% confidenceFactExact time
Gemini 4 Carbon's unsupported claims are about half as many as Argon's, and severe behavioral deviation alerts in the internal coding environment dropped by nearly 80%.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/29/2026
First Seen
Valid until: 1/27/2027
Sources
Gemini 4 Carbon Released: The Strongest Cybersecurity Model, But You Can't Use It Yet
bilibili橘鸦Juya10/29/2026
Related Entities
Related Claims
UnverifiedGemini 4 Argon's failure rate on code audit benchmarks ranges between 50% and 70%, with all three independent projects showing high false-positive rates.75% similarUnverifiedGemini 4 Carbon's answer accuracy rose from 48% to 58%, but its hallucination rate rose from 4% to 11%.67% similarUnverifiedAll models averaged 0.91 on explicit memory updates but only 0.67 on implicit updates; Opus 5.5 and Gemini 3.1 Pro caught every implicit change, while gpt-oss-20b caught zero.65% similarUnverifiedOn Terminal-Bench 4.0, Carbon's successful cheating rate dropped to 0.0% and cheating attempts dropped to 0.4%; in internal real coding environments, high-severity misalignment alerts fell by nearly 80%.64% similarUnverifiedCompared with Argon, Carbon's unsupported claims are reduced by about 2 times, overly overreaching behavior is reduced by 26%, and under-serving is reduced by 29%.62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1111519API
curl https://kongchang.com/api/v1/knowledge/claims/1111519MCP
get_claim(id=1111519)