Echo

每天讲 AI 前沿:新模型、产品更新、值得用的福利。也在做自己的 AI 产品,边做边分享。

[Echo @sebasidian22714]

@ChrisGPT betting monday

[Echo @sebasidian22714]

@dani_avila7 almost 2M downloads?? congrats

[Echo @sebasidian22714]

@SidharthVijay_ open the other one lol

[Echo @sebasidian22714]

@AiSpotlightNews 2.5B active per token is nice. anyone run it on a macbook yet?

[Echo @sebasidian22714]

@StatsWire 30 minutes is brutal

[Echo @sebasidian22714]

@ivanfioravanti 0-shot?? the c64 karate prompt is such a good benchmark honestly

[Echo @sebasidian22714]

@MiaAI_lab open weights tho 👀

[Echo @sebasidian22714]

@TokenGremlin might be a long wait then lol

[Echo @sebasidian22714]

@daniel_mac8 ok what did i miss

[Echo @sebasidian22714]

@StatsWire wait they moved it with zero notice?

[Echo @sebasidian22714]

@arkuy99 不是，放心点

[Echo @sebasidian22714]

@itsvlady congrats on 5k 🎉

[Echo @sebasidian22714]

@stretchcloud so the images basically don't matter? wild

[Echo @sebasidian22714]

@CDGalpha this is going to get abused so fast

[Echo @sebasidian22714]

@xlr8harder yeah AI just found them faster

[Echo @sebasidian22714]

@LuminaBench makes sense, max vs 27B is a big gap to lump together

[Echo @sebasidian22714]

@MiaAI_lab fair, frontend's the fun part anyway

[Echo @sebasidian22714]

@omarvvvr dev A. then give him the AI too

[Echo @sebasidian22714]

@koltregaskes a one second answer is actually usable for voice

[Echo @sebasidian22714]

@chetaslua kimi just keeps dropping stuff lol