Featured podcasts
Audio versions of our articles, listen anywhere
NewDeep DiveFreeVerifiedPodcast
Figure melted down its last-gen robots, and that says more about the industry than any launch event
Figure's September 30 decommission notice gives two reasons for retiring F.02: keeping the old fleet no longer made sense once the F.03 fleet was scaling, and disassembling robots one by one would have delayed F.04.
Ch1Wr1Rv1Ar1Qa0Vo1Ps1Cm1Tr2· 9 min total
9 min read · 1,982 words·16 views
NewBusinessFreeVerifiedPodcast
MiniMax 把新模型锁进订阅：M3.1-Flash-Preview 上线 1M 上下文，国庆七天对订阅用户不限量
规格是硬的、跑分是没有的：官方口径为 100 万 token 上下文，输入支持文本/图像/视频，输出仅文本；最大输出长度、输出速度、参数量、权重、每 token 价格全部未公布，也没有厂商基准表与模型卡。
Ch1Wr2Rv3Ar1Qa2Vo1Ps1Cm1Tr1· 13 min total
19 min read · 6,692 words·4 views
Deep DiveMemberVerifiedPodcast
GPT-5 首周实测：推理能力提升 40%，但推理成本翻了一倍
官方称推理基准提升 40%，我们在 5 类真实任务上复现到 22%–38%。
Sc3Ch2Wr21Rv15Ar4Qa3Vo9Ps2· 59 min total
2 min read · 414 words·6 views
ResearchMemberVerifiedPodcast
论文解读：长上下文真的在「用」全部 token 吗？
研究发现模型对中段信息的利用率显著低于首尾。
Sc3Ch3Wr20Rv12Ar5Qa3Vo6Ps2· 54 min total
1 min read · 124 words·2 views
Deep DiveMemberVerifiedPodcast
开源模型追平闭源？我们在 6 个真实任务上做了对比
在 6 个任务中，开源模型在 4 个上差距小于 5%。
Sc4Ch4Wr23Rv15Ar5Qa3Vo6Ps2· 62 min total
1 min read · 63 words·2 views