Sean Highness

[Sean Highness @seanhighness]

@mikeydsoftware Hahahah

[Sean Highness @seanhighness]

Burned my codex limit from a cruise ship with no internet , new achievement unlocked

[Sean Highness @seanhighness]

@0xWhiteMage What runtimes support exl3 and what about speculative decode

[Sean Highness @seanhighness]

@mr_r0b0t @NVIDIAAI @JensenHuang @huggingface Amazing

[Sean Highness @seanhighness]

@JensenHuang @huggingface Let’s go

[Sean Highness @seanhighness]

@tonysimons_ Love to see it

[Sean Highness @seanhighness]

@FruthTalse @witcheer My build gives 325k token pool with dflash2 on a single 5090

[Sean Highness @seanhighness]

@tonysimons_ Was just asking for this

[Sean Highness @seanhighness]

@loktar00 Wild ! Good to see people testing themselves against assumptions

[Sean Highness @seanhighness]

@jun_song This is epic

[Sean Highness @seanhighness]

@ItsmeAjayKV Congrats

[Sean Highness @seanhighness]

@loktar00 Got one in socal, might wait till they are more expensive lol

[Sean Highness @seanhighness]

@superalesha @QwenDevs 27b for Eva

[Sean Highness @seanhighness]

@Blackwellboy All the open models working together create unlimited abundance in 90 days

[Sean Highness @seanhighness]

@loktar00 The kv density in nice too

[Sean Highness @seanhighness]

@tonysimons_ Let’s go

[Sean Highness @seanhighness]

Escha W2 at 3666 t/s prefill in beellama on 5090 surpasses native runtime. Time to fix decode and then take 12gb down to 8.6 with same quality - get the 3060’s and 4070’s ready

[Sean Highness @seanhighness]

@tonysimons_ Slopen Claw

[Sean Highness @seanhighness]

@MiaAI_lab Amazing what runtime?

you can do high concurrency in vllm with this https://t.co/B1Wzm5OY6o 325k total kv pool- 5090 only tho :) 200 t/s

[Sean Highness @seanhighness]

Is back up for me

[Sean Highness @seanhighness]

GPT work is DOWN ?