David solomon

Cooking up something new | ex: Founder @nreach_io | AI Researcher

[David solomon @soldi_sol]

@Hikari_07_jp That's slower than rtx6k LiL repo

[David solomon @soldi_sol]

The next cliff is liquid cooling.

[David solomon @soldi_sol]

@0xIlyy He's running a "count to 1000" prompt with 1000 spec decoded tokens.

BS

[David solomon @soldi_sol]

@superalesha @TheAhmadOsman its great news :)

[David solomon @soldi_sol]

@superalesha I'd love to make it as a new compaction method on omp's setting.
the "state diff" is interesting and logical. git - like, for sure faster

[David solomon @soldi_sol]

@superalesha testing it out as we speak!

[David solomon @soldi_sol]

@superalesha @TheAhmadOsman omfg

[David solomon @soldi_sol]

@superalesha @TheAhmadOsman cool waiting for it!

[David solomon @soldi_sol]

@superalesha @TheAhmadOsman Can you share it?

[David solomon @soldi_sol]

@1337hero interesting! love it, its smart

[David solomon @soldi_sol]

@1337hero what does this verify param means ? what's the context here

[David solomon @soldi_sol]

@1337hero where's that model then? i wanna try it. regular deepseek 0731 doesnt go above glm in most cases, his reasoning is lacking many times. can you share the model weights

[David solomon @soldi_sol]

@0xSero thanks for running the benchmark! have been waiting for this

[David solomon @soldi_sol]

@miao_xiong_cs we're an applied research lab, still in stealth. interested?

[David solomon @soldi_sol]

@MiaAI_lab @baseten yep. on the baseten version. might be a very bad quantized model..

[David solomon @soldi_sol]

@SebastiaanDuijn @AtharvaIngle7 Can you share your PR here? Will check it put

[David solomon @soldi_sol]

Kimi is a superior product. period.
its better on cost, speed, and same intelligence

@Kimi_Moonshot