Ganesh Babu

Building Enterprise Supply Chain Agents

[Ganesh Babu @appakaradi]

@sama Is that some kind of code word to what is coming ?

[Ganesh Babu @appakaradi]

@osanseviero Better reasoning, better tool calling , better than Qwen 3.6 27B and 35 A3B.

[Ganesh Babu @appakaradi]

@pashmerepat Karma is real.

[Ganesh Babu @appakaradi]

@tobi @nvidia Can I have one please?

[Ganesh Babu @appakaradi]

@awnihannun 1 billion is peanuts for apple when it comes to risk mitigation. Also, it is fair to pay for the model even though they have similar option available for free

[Ganesh Babu @appakaradi]

@LiuYunlong63318 They are working on it.. Give enough compute and data, with some good team, Microsoft is very capable of building near state of the art models.

[Ganesh Babu @appakaradi]

@YanhongLi2062 @thinkymachines @allen_ai @lambdaviking @soldni @allen_ai can i be @YanhongLi2062 's replacement?

[Ganesh Babu @appakaradi]

@loktar00 @SpaceTimeViking Then you should write clearly that you are hoping. Your post sounds like you are infering some signals.

[Ganesh Babu @appakaradi]

@Alibaba_Qwen @rafaelcaricio 27b please

[Ganesh Babu @appakaradi]

@AnthropicAI what good are you when you can not help in the moment of need? I asked you to review my father’s medical records and treatment who is battling a cancer. Your response is switched to Opus. What good is Fable if you can help in the hour of the need?

[Ganesh Babu @appakaradi]

@LottoLabs I do not see any hope for smaller models like 27B and 35 A3B. They just want to compete at the top level.

[Ganesh Babu @appakaradi]

@HermesAgentTips Qwen 3.6 27B.

[Ganesh Babu @appakaradi]

@thsottiaux If my company unblocks open ai and codex , that will bring me joy.

[Ganesh Babu @appakaradi]

@trq212 What is the hard part? You already have model chdckpoint. You just have to scale the inference assuming you have the hardware. What is so hard about this unless you are introducing optimization on the inference stack

[Ganesh Babu @appakaradi]

@ivanfioravanti yes.. codex, cursor, anti gravity - flash is fast, open code kimi and deep seek, pi - qwen 3.6 27B, claude code - fable ans opus

[Ganesh Babu @appakaradi]

@triketora @AnthropicAI in what way?

[Ganesh Babu @appakaradi]

@mweinbach This is incredible!!

[Ganesh Babu @appakaradi]

@chrisgpt is it API only? or are we getting 27B and 35A3b too?

[Ganesh Babu @appakaradi]

GPT 5.6 Sol's ability to reason complex problems and all the nuances is incredible. It cleans up everything Opus missed. Codex harness and UI is a big relief from multi terminal hell (even with cmux, orca,etc.etc)..

[Ganesh Babu @appakaradi]

@OfficialLoganK can we make the AI Studio use normal enter instead of Control Enter to submit messages?