360 Tulongfeng tops CyberGym with a 99.4% vulnerability reproduction success rate

1. 13:00
   WebAI Threat Watch (aithreat.io)· Media coverage

   99.4% tops the world at first place! This Chinese AI, trained by a world champion(opens the original post in a new tab)

1. 18:44
   X@Techub_News· Discussion

   🏆 360's Tulong Feng tops the CyberGym AI security leaderboard with a 99.4% success rate 360's Tulong Feng tops CyberGym AI security leaderboard with 99.4% success rat…(opens the original post in a new tab)
2. 17:51
   X@shanfengfengybq· Discussion

   A 99.4% vulnerability reproduction rate is scarier than topping the coding leaderboard: Agent swarms have turned attack and defense into an assembly line. Previously people stayed up late scanning CVEs; now it's about who can feed real vulnerabilities into agent clusters first. Agents that can write code are everywhere; only one that can reproduce a real vulnerability and break through you is what counts as productivity. #AI #Agent https://…(opens the original post in a new tab)
3. 17:50
   X@swordBluesy· Discussion

   The number 99.4% is indeed solid—1507 real vulnerabilities held above 99% across three rounds, so it's not a fluke. But your own reminder matters more: this is only a reproduction score under a specific benchmark, not the same as winning every real attack and defense; being first on the leaderboard and actually being able to find vulnerabilities are two different things, and the 15,000 vulnerabilities already found are more convincing than the ranking. By the way, …(opens the original post in a new tab)
4. 17:48
   X@knightleijun· Discussion

   99.4% vulnerability reproduction rate, 360's Tulong Feng tops the AI security "World Cup" CyberGym. 1507 real vulnerabilities, three rounds at 99.27%/99.47%/99.40%, mean 99.38%, setting a new record; the same leaderboard also includes OpenAI, Anthropic, Goog…(opens the original post in a new tab)
5. 17:15
   X@NFT_Chen· Discussion

   💥Didn't expect this! 360's self-developed AI attack-defense system Tulongfeng topped the global ranking in the AI Security World Cup CyberGym with a 99.4% vulnerability reproduction success rate, crushing GPT-5.5-Cyber, Mythos Preview and Gemini 3.8 Flash Cyber! 🔹Three rounds of evaluation averaged 9…(opens the original post in a new tab)
6. 13:19
   X@ppmmzzcn· Discussion

   🤖[QbitAI] 99.4% tops the global ranking! This Chinese AI learned from a world champion. On the CyberGym leaderboard known as the "AI Security World Cup," 360's self-developed AI security attack-defense system Tulongfeng topped the ranking with a 99.4% vulnerability reproduction success rate, setting the highest record on the leaderboard. In 2020, 360 Alpha…(opens the original post in a new tab)
7. 13:12
   XiaohongshuXiaohongshu · Tech Just Now· Discussion

   360's Tulongfeng tops the world at 99.4%(opens the original post in a new tab)
8. 13:01
   WebQbitAI· Media coverage

   99.4% tops the world at first place! This Chinese AI, trained by a world champion(opens the original post in a new tab)
9. 12:00
   WeChat · QbitAI

   360 Tulongfeng tops CyberGym globally with a 99.4% vulnerability reproduction success rate, descending from the Pwn2Own world championship team

   360's self-developed AI security attack-and-defense system Tulongfeng topped the CyberGym leaderboard globally with a 99.4% vulnerability reproduction success rate. It descends from the 360 security team that once won the Pwn2Own world championship, and is powered by the collaboration of security-domain large models and specialized agent swarms; across three rounds of independent evaluation on 1507 real vulnerability tasks, its success rate exceeded 99% in all rounds. Since its release in June this year, Tulongfeng has cumulatively discovered more than 15,000 vulnerabilities, and over 500 enterprises have onboarded for testing.

1. 18:52
   WeChatDoonsec· Media coverageIn Chinese, as written

   世界第一！360图龙锋登顶CyberGym国际AI安全榜单(opens the original post in a new tab)
2. 18:26
   WeChatDoonsec· Media coverageIn Chinese, as written

   世界第一！360图龙锋登顶CyberGym国际AI安全榜单(opens the original post in a new tab)
3. Time not given
   Sohu

   360's self-developed AI security attack-defense system Tulongfeng tops CyberGym with a 99.4% vulnerability reproduction success rate

   360's self-developed AI security attack-defense system Túlóngfēng topped CyberGym with a 99.4% vulnerability reproduction success rate and set a new record; the same leaderboard includes models and agents from OpenAI, Anthropic, and Google. The system underwent three rounds of independent evaluation across 1507 tasks, with an average success rate of 99.38%, adopting a "security-domain large model + specialized agent swarm" approach. It has cumulatively discovered over 15000 vulnerabilities, and more than 500 enterprises have onboarded.

Current attention 19·Peak in the comparable period 19 (Nov 6 03:00)·Change over the comparable past 24 hours –