TechMD

🔬 MD fueled by a deep passion for Medicine & Tech. 🌐 Exploring the frontiers of VR/AR/MR & BCI. 🤖 AI enthusiast.

[TechMD @TechMDAI]

The future may belong not to those who know the most, but to those whose identities can survive being transformed by what they learn.

[TechMD @TechMDAI]

@bitale14 @YonasMeshesha @MiaAI_lab Congratulations

[TechMD @TechMDAI]

I think DeepSeek V4.1 Flash is raising the cost of the RTX Pro 6000. I’m seeing 500–600 tk/s with that model.

[TechMD @TechMDAI]

@YonasMeshesha @bitale14 @MiaAI_lab Thank you

[TechMD @TechMDAI]

@steddyman It’s a blessing and I’m humbled

[TechMD @TechMDAI]

It’s crazy I’m using my 4 DGX sparks more than my 2 RTX pro 6000 currently. What world are we living in.

[TechMD @TechMDAI]

@CalebBaskin Yes

[TechMD @TechMDAI]

Got this running on 2 Pro 6000s at 140 tk/s
https://t.co/Zfjn30YHYi

[TechMD @TechMDAI]

@tfolkman Yes but I still use the pros coding with glm flash

[Jim McPherson @JamesMcPherson]

The model that makes $7000 Sparks a good value.
If you are running 4 Sparks or more, this is why you bought them.
DeepSeek 4.1 Flash 1017
* Near frontier intelligence.
* 120+ tok/sec on code
* 80+ tok/sec on prose single stream.
* 285 tok/sec with 4 agents writing code.
* Vision
* 1 Million token context window
* RAM to spare, you won’t be fighting OOM on every context window.

[TechMD @TechMDAI]

@krunkosaurus Me station

[TechMD @TechMDAI]

Congratulations

[Espen @EspenJD]

Good news! RMA approved, I'm getting a replacement.

Funny part: they approved it literally 3 minutes after confirming they'd received the card lol

[TechMD @TechMDAI]

@krisg7gr63 Not really a GR fan, but have respect for him. This car has none of his DNA. It is definitely built for Kimi’sstyle. I’m not sure, but over the last couple of years, Mercedes doesn’t like driver input.

[TechMD @TechMDAI]

@Nitin_wysiwyg They have 4 rtx pro 6000. Its rips with 4

[Yash Khurana @yolo_yash_]

@djnmrjnvc @TechMDAI Yeah, just couldn't find a recipe for it. Made one myself in case anyone else needs it - https://t.co/AWL4YtQp5U (uses diffbots weights)

[TechMD @TechMDAI]

@gabu3d_pl I need to try maya honestly

[TechMD @TechMDAI]

@squarrier @dangerm00se @wrldsuksgo2mars I need this in my life

[TechMD @TechMDAI]

Deepseek v4.1 on 4 DGX sparks is goated, wow just wow 🤯

[Stephen J Ronan MD @SJRonanMD]

@siegl_rene @somora_ai @openclaw @NousResearch @LiteLLM Two DGX Stations, each with an RTX PRO 6000 as an expert sidecar, plus three Sparks for small and specialist models, all behind one gateway like yours. What surprised me: speed stopped being the hard part months ago. Now it's deciding which model gets which job.

[Kalshi @Kalshi]

JUST IN: Anthropic to release S-1 IPO filing at 9 AM EST

[Niklas @niklaslenz_ai]

DeepSeek-V4.1-Flash-1017 decodes code at 80.5 tok/s on one Mac Studio M5 Ultra 256 GB with BobClawblaw's recipe on TensorFold.

Setup:
– Recipe and TensorFold branch deepseek-v41-main (72a552ed), his own tool and settings
– Context 524,288 tokens, 4 streams, DSpark drafts on, temperature 0
– Speed and quality with thinking off, except GPQA with thinking

Speed:
– Decode, code: 80.5 tok/s
– Decode, prose: 70.5 tok/s
– 500 words: about 9.3 s
– First token after a 32K prompt: 18.6 s
– 8 requests at once: 88.3 tok/s in total

Quality:
– MMLU-Pro: 67.5% ± 4.6 points
– GPQA Diamond: 45.0% ± 6.9 points
– IFEval: 90.5% ± 4.1 points
– LiveCodeBench: 60.0% ± 14.5 points
– GPQA with thinking: 76.7% ± 14.6 points

Voxel world:
– Built as an agent in 45.7 min, DeepSeek Harness 0.2.1-rc.2, one prompt, one try, no help, thinking on
– Tokens written: 112,065
– Browser check: 8.5/10

Good to know:
– The author runs temperature 0, the DeepSeek card suggests 1.0. I kept the author's setting.
– The browser check ran on a Linux box from the same files.
– Answers cut at the cap count as wrong: LiveCodeBench 14 of 40, GPQA Diamond 9 of 198, MMLU-Pro 1 of 400.

Play the world in your browser. All numbers, the exact command and the raw output:
https://t.co/DXdZ0u8V2e

Recipe by @BobClawblaw on @ashxhart's TensorFold, model by @deepseek_ai.

[TechMD @TechMDAI]

@FundedXX @bijanbowen Still use them

[TechMD @TechMDAI]

@lalopenguin lol I still use em.

[TechMD @TechMDAI]

@XOdJE16179 lol 😂 yup that’s all what counts