𝗭𝗲𝗻 𝗠𝗮𝗴𝗻𝗲𝘁𝘀

Magnets must be respected, but need not be feared. 𝕏 profile subsumed by Shihan Qu, founder of ᐅfed bannedᐊ Zen Magnets. Micromags still @

[𝗭𝗲𝗻 𝗠𝗮𝗴𝗻𝗲𝘁𝘀 @ZenMagnets]

Mohit's Imajev now in FP8 quant and optimized for SM120. 0.1% accuracy loss from the BF16 version and ~50% faster. Should work well for SM121 too.
https://t.co/j1IL4fnVP6

[Mohit Garg @mohitt_garg]

4 days ago I launched Imajev.

This week the official board ranked it
#1 of 91 on JevBench (v1.4.2.2, scored 27 Sep) - Ahead of Jev 1.13.0 itself.

#3 of 56 on DecisionBench, above GPT-5.6 Luna and a 552B DeepSeek

[Lex Fridman @lexfridman]

NVIDIA letting Jensen go is a huge mistake, for America and for humanity. It's Apple firing Steve Jobs all over again, but much bigger, given the role of AI in defining the future of our world.

I've been keeping quiet on this out of respect for Jensen's family and team, who I've met and who are amazing humans, and because I personally hate infidelity. So my heart goes out to them in this difficult personal family moment.

But I can also say that Jensen is human, and makes mistakes like all of us do. Not that I matter at all, but I've known Jensen for a long time and have gotten the chance to spend a lot of time with him, behind the scenes, away from mics and cameras. I have seen him show kindness, humility, care to everyone around him, including strangers, including from the poorest most struggling parts of society, treating all the same with the same thoughtfulness and attention, because they are fellow humans and not because he has something to gain from them. That shows rare character, to maintain humility and humanity through all of it.

Another data point is what people who know him well have said privately to me about Jensen over the years. I've never had anyone tell me anything but how awesome he is or how much they love him. Probably over a thousand people at this point. This includes hundreds of engineers, CEOs, colleagues, friends, etc.

Also, since there was some spy talk. Jensen is as proud an American as it gets. I'm in shock frankly at all of this, including the mass hysteria around it. I love America, and I think NVIDIA letting Jensen go is terrible for America.

I stand with Jensen.

Love you all ❤️

[𝗭𝗲𝗻 𝗠𝗮𝗴𝗻𝗲𝘁𝘀 @ZenMagnets]

@liquidtrading If only that were the timeline we were on

[𝗭𝗲𝗻 𝗠𝗮𝗴𝗻𝗲𝘁𝘀 @ZenMagnets]

@jjstylez315 @liquidtrading Whups

[𝗭𝗲𝗻 𝗠𝗮𝗴𝗻𝗲𝘁𝘀 @ZenMagnets]

Bad choice nvidia. Shoulda kept the goat at the helm.

[𝗭𝗲𝗻 𝗠𝗮𝗴𝗻𝗲𝘁𝘀 @ZenMagnets]

AI boom/bubble popped by a trillion dollar dick. Man shoulda just leaned into it. Changed signature to "Yeah, I fuck" and signed a few more breasts.

[The Kobeissi Letter @KobeissiLetter]

BREAKING: Nvidia stock, $NVDA, plunges over -10% after Jensen Huang officially resigns as CEO.

The stock has now erased -$750 billion in market cap since the scandal began.

[𝗭𝗲𝗻 𝗠𝗮𝗴𝗻𝗲𝘁𝘀 @ZenMagnets]

The recipe I've been main-ing
https://t.co/ajEAPOQ7w8

[𝗭𝗲𝗻 𝗠𝗮𝗴𝗻𝗲𝘁𝘀 @ZenMagnets]

Benjamin is exactly right. Tok/s speeds are meaningless when we're not measuring on the same track.

These days I've been using llm-inference-bench as my metric for optimization, with 10x runs for all C1 samples.

Deepseek flash 1017: Currently getting ~365-335 tok/s C1, 2200 tok/s C32, with a fused D4 hybrid of brrr and hyperflare, with 16M KV available. 4x rtx6k 450w, ram engrams, and direct to mobo pcie connection.

[Benjamin Marie @bnjmn_marie]

Feels like “tok/sec” is losing its meaning here.

I keep seeing numbers that are nowhere near achievable in real-world tasks.

I get that X’s algorithm rewards this kind of hype, but it’s a shame to see some genuinely great custom inference engines putting their reputations at risk just to chase impressive numbers.

[𝗭𝗲𝗻 𝗠𝗮𝗴𝗻𝗲𝘁𝘀 @ZenMagnets]

@keennay If you feel like an experiment, I think this recipe can achieve faster with your hardware, without many sacrifices

[𝗭𝗲𝗻 𝗠𝗮𝗴𝗻𝗲𝘁𝘀 @ZenMagnets]

Benjamin is exactly right. Tok/s speeds are meaningless when we're not measuring on the same track.

These days I've been using llm-inference-bench as my metric for optimization, with 10x runs for all C1 samples.

Deepseek flash 1017: Currently getting ~365-335 tok/s C1, 2200 tok/s C32, with a fused D4 hybrid of brrr and hyperflare, with 16M KV available. 4x rtx6k 450w, ram engrams, and direct to mobo pcie connection.

[𝗭𝗲𝗻 𝗠𝗮𝗴𝗻𝗲𝘁𝘀 @ZenMagnets]

@calboDev Company still exists. For the first time in like ten years, we're going to introduce a new product (sometime this month). It's not magnetic though

[𝗭𝗲𝗻 𝗠𝗮𝗴𝗻𝗲𝘁𝘀 @ZenMagnets]

@itsjustmarky 700 tok/s measured how?

[𝗭𝗲𝗻 𝗠𝗮𝗴𝗻𝗲𝘁𝘀 @ZenMagnets]

@OrangeBubb43389 @omarvvvr If that were true, we'd not use gpus

[𝗭𝗲𝗻 𝗠𝗮𝗴𝗻𝗲𝘁𝘀 @ZenMagnets]

@MiaAI_lab Api works too

[𝗭𝗲𝗻 𝗠𝗮𝗴𝗻𝗲𝘁𝘀 @ZenMagnets]

@Khan_Fatimaa1 @grok is there evidence for this claim?

[𝗭𝗲𝗻 𝗠𝗮𝗴𝗻𝗲𝘁𝘀 @ZenMagnets]

@AlecLace @grok Where can I find the full Dave Chappelle clip with context about what he said

[𝗭𝗲𝗻 𝗠𝗮𝗴𝗻𝗲𝘁𝘀 @ZenMagnets]

@AlecLace Someone please post THE FULL CONTEXT

[𝗭𝗲𝗻 𝗠𝗮𝗴𝗻𝗲𝘁𝘀 @ZenMagnets]

@FreeThinkerFit Need the full context

[𝗭𝗲𝗻 𝗠𝗮𝗴𝗻𝗲𝘁𝘀 @ZenMagnets]

@LottoLabs High probability Jensen was honeydicking fake info straight to the CCP, Iyam

[𝗭𝗲𝗻 𝗠𝗮𝗴𝗻𝗲𝘁𝘀 @ZenMagnets]

@0xSero @LLMJunky Get a room nerdssss

[𝗭𝗲𝗻 𝗠𝗮𝗴𝗻𝗲𝘁𝘀 @ZenMagnets]

@LLMJunky

[𝗭𝗲𝗻 𝗠𝗮𝗴𝗻𝗲𝘁𝘀 @ZenMagnets]

@ltgiv @grok who did it