Ismail Elsherbini

Co-Founder @ VERSO

[Ismail Elsherbini @elsherbin_]

@parkerinai @deepseek_ai What's the difference to you, the literally named the previus version with a date and model name, you are complaining that they used 4.1🤣🤣🤣

[Ismail Elsherbini @elsherbin_]

@0xSero Why not 64gb running Qwen 3.8 flash next using DwarfStar configuration?

[Ismail Elsherbini @elsherbin_]

@davis7 Bro muse spark 1.3 contributor tier is way cheaper and better than Luna and glm 5.3 like why did you put it so low lol

[Ismail Elsherbini @elsherbin_]

@forgebitz Use pi or t3code

[Ismail Elsherbini @elsherbin_]

@imnotchalk @opencode @thdxr Yes , I realized it's glm, lol

[Ismail Elsherbini @elsherbin_]

@alexgetmancom Astra 6 Pro? Not the base version ?

[Ismail Elsherbini @elsherbin_]

@XFreeze I think AA isnt able to keep up with ai advancements bro

[Ismail Elsherbini @elsherbin_]

@imnotchalk @opencode @thdxr Probably Xiaomi, we will see the encoder , they have a model coming up. Mimo v3 , also the fact that they are testing it on opencode go with limited usage , might be an indicator it's a big model

[Ismail Elsherbini @elsherbin_]

@eidzoku 65

[Ismail Elsherbini @elsherbin_]

@DavidOndrej1 @jukan05 Gemini 3.8 + muse spark 1.3 == astra

[Ismail Elsherbini @elsherbin_]

@SSHCodes I need a benchamarks benchamark benchamks , when models start benchmaxing the benchmax benchmark

[Ismail Elsherbini @elsherbin_]

@Takumixbt @ChangJonathanC Trust me , its cheaper

[Ismail Elsherbini @elsherbin_]

@jetpackjoe_ @theo @maria_rcks It scored worse on deepswe , but cheaper at 67.5%

[Ismail Elsherbini @elsherbin_]

@robinebers @devindesktop Why don't they update the checkpoint to glm 5.3 bruv

[Ismail Elsherbini @elsherbin_]

@scaling01 It also got worse in deepswe ? These benchamarks are getting weird , how are Chinese models and Luna outperforming fable 5.1

[Ismail Elsherbini @elsherbin_]

@MatthewBerman @bot That's awesome , I hope you can help me try it out 🙏

[Ismail Elsherbini @elsherbin_]

@fabianstelzer Minimax h3 max by fal ai , crushes this

[Ismail Elsherbini @elsherbin_]

@theo Multimodal you can use ox Alpha , it's free and will be probably cheap glm flash, A tier model

[Ismail Elsherbini @elsherbin_]

@abhijitwt He is ranking based on multiple factors not intelligence alone , but cost and speed also , he doesn't like terra because it's expensive compared to Luna which is basically free and there isn't a big gap in infeligence , you can have sol orchestrate Luna sub agents

[Naval @naval]

Immortality, meaning, and love are readily available - they’re just running around disguised as children.