Tibo reveals OpenAI's second internal model, Guardian
Relaying Tibo's claims: Computer Use got 10x faster in a year through harness plus model changes, Astra was used to optimize parallelism and latency, and a second internal model, Guardian, watches over the main agent.
Tibo 首次曝光第二套内部模型 Guardian,并撂下狠话! 并亲口说了句很猛的话: Computer Use 一年提速 10 倍,不是模型吐字变快,是 harness + 模型一起改。 内部还有第二套模型叫 Guardian,专门盯着主 Agent。 Astra 被拿来大规模优化并行、砍卡顿、压缩“想多久才能点下一步”的预算。 他把这些叠在一起,才做出 10x。 这已经不是推理速度竞赛了,是内部模型在互相盯梢、互相加速。 #OpenAI #Tibo #Astra #Guardian #ComputerUse #AIAgent #DevDay https://x.com/nft_chen/status/2105180098722947236?s=46 https://x.com/tbpn/status/2105084249418486059/video/1
This post quotes another post: x.com/nft_chen/status/2105180098722947236