Unverified50% confidenceBenchmarkExact time
In tests such as GDP and PDF, Gemini 4 Hydrogen ties for first place with a hallucination rate of 8%, while under the same evaluation Sonnet 5.5 is at 43% and GPT-6.1 Sol at 52%.
1
Sources
50%
Confidence
Short-term (~14 days)
Relevance
11/14/2026
First Seen
Valid until: 11/28/2026
Sources
Gemini 4 Hydrogen Released: Hands-On Testing of the Fastest, Cheapest Flash Model
bilibiliAI奥德赛11/14/2026
Related Entities
Related Claims
UnverifiedGemini 4 Hydrogen scored 57% on the Terminal Bench 4.1 test, more than three times the previous generation.77% similarUnverifiedFabrication rates when the total is fully obscured: Claude Sonnet 5 at 38%, Gemini 3 Flash at 21%, Gemini 3.1 Flash-Lite at 44%, GPT-5.4 nano at 73%; fabrication rates when the total is half-obscured: Claude at 92%, the other three models all at 100%.70% similarUnverifiedGemini 4 Hydrogen scores 38.2% on the FrontierCode 2.0 benchmark.70% similarUnverifiedOn the Artificial Analysis Intelligence Index, Gemini 4 Hydrogen scores 51 at the default high thinking level and can reach 54 at the highest xhigh thinking setting.68% similarUnverified2024-2025年间,Claude 3.5 Sonnet、GPT-4o、Gemini 2.0等模型在HumanEval、SWE-bench等编程基准测试的通过率从不足50%提升至90%以上67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1171317API
curl https://kongchang.com/api/v1/knowledge/claims/1171317MCP
get_claim(id=1171317)