Dataphile

Building sovereign SI infrastructure. 15-node cluster · 6× RTX PRO 6000 Blackwell (576GB VRAM) · 2TB ECC RAM · PB-scale Ceph · 200GbE fabric

[Dataphile @Dataphil3]

MiniMax-M3.1-Flash, day 0, running local.

212.6 tok/s, batch 1, verified on LocalMaxxing.

428B MoE, NVFP4 experts, DSpark speculative decoding. vLLM day-0 image, TP4 on one workstation.

https://t.co/EKAHy6M42I

#MiniMax #LocalLLM

[Dataphile @Dataphil3]

@VaibhavLab mythos is only for a few selected people

[Dataphile @Dataphil3]

@komal_uk01 he is a business man

[Dataphile @Dataphil3]

@Greg_GL_87 @xodud_rkd 600gb isnt so bad

[Dataphile @Dataphil3]

@shadownet1337 @deepseek_ai @Google yes they set the stage for the AI era, yet they didnt break out the chatbots and have been lagging on the frontier, but are making a big comeback now. they have all the ingredients for major Ai dominane

[Dataphile @Dataphil3]

Same box. Same model. Code: 502 tok/s. Prose: 211.

@deepseek_ai V4.1-Flash-1017, batch 1, one workstation:
code lands 6.1 tokens per step (76% of drafts accepted)
prose lands 2.1 (39%)
step time: 12 vs 10 ms

Spec decoding benchmarks measure your prompt.

#DeepSeek #LocalLLM

[Dataphile @Dataphil3]

@alwayspriyesh Great for deep review

[Dataphile @Dataphil3]

@shadownet1337 @deepseek_ai Remarkable how much @google has contributed to Ai research over the years

[Dataphile @Dataphil3]

@yangyang_2048 @grok Is he Ai ?

[Dataphile @Dataphil3]

@shadownet1337 @deepseek_ai Google is lid

[Dataphile @Dataphil3]

@blessedofcodes Toaster

[Dataphile @Dataphil3]

@pfernan95dev 0.42$

[Dataphile @Dataphil3]

@Gaus450 where in USA? where in Australia?

[Dataphile @Dataphil3]

@PeasantSmith @Blackfrost_AI Splendid

[Dataphile @Dataphil3]

@Blackwellboy Glorious

[Dataphile @Dataphil3]

@Blackwellboy Glorious !

[Dataphile @Dataphil3]

@markfenner that mug again! I want it in film negative, black with white, instead of white on black.

[Dataphile @Dataphil3]

@_jayyy18_13k pro means TOOBIG for my infra 2T parameters prob

[Dataphile @Dataphil3]

@theo But it’s NOT in the bloody APi

[Dataphile @Dataphil3]

@MiaAI_lab https://t.co/CyXjku90NA

[Dataphile @Dataphil3]

MiniMax-M3.1-Flash, day 0, running local.

212.6 tok/s, batch 1, verified on LocalMaxxing.

428B MoE, NVFP4 experts, DSpark speculative decoding. vLLM day-0 image, TP4 on one workstation.

https://t.co/EKAHy6M42I

#MiniMax #LocalLLM

[Dataphile @Dataphil3]

@LLMJunky thats true, yes

[Dataphile @Dataphil3]

@karbonx6 We will soon all be #Eloi due to @claudeai