Pinned Tweet
These guys are rockstars - we are delighted to be powering their model training at @verdacloud
Downscale medical imaging to fit in GPU memory and the model loses the detail that catches faint findings.
@epsilon_hq built its next-gen radiology AI models on a dedicated NVIDIA Blackwell Ultra cluster on Verda, now running 5x more experiments/week and training 3x faster.
Full story: verda.com/blog/epsilon-healt…
Andy Long retweeted
The @epsilon_hq team is proud to partner with @verdacloud to accelerate our radiology AI research.
Thanks to Verda, we've been able to train our models on dedicated @nvidia Blackwell Ultra GPUs, allowing our research team to do 5x more experiments per week and significantly faster training.
Moreover, with Verda’s infrastructure and engineering support, our research team can spend less time managing infrastructure and more time advancing research.
Above all, helping save more lives.
Huge thanks to the Verda team.
@arjunkarpur @aakamishra1
#nvidia #healthcare #infrastructure #AI #compute
Downscale medical imaging to fit in GPU memory and the model loses the detail that catches faint findings.
@epsilon_hq built its next-gen radiology AI models on a dedicated NVIDIA Blackwell Ultra cluster on Verda, now running 5x more experiments/week and training 3x faster.
Full story: verda.com/blog/epsilon-healt…
Very true Eli. You need more than MW and $$$.
Why @verdacloud is 🚀
to be a real neocloud you need to build software, not just supply power and GPUs
lots of people are buying sites and filling them with B300s
that’s cool, they’ll probably sell a few racks, but they’re gonna get blown out of the water by companies that build software
spot is fundamentally a software product. the platform is constantly shifting workloads around and cleaning up between jobs, and you’re solving many orchestration problems at scale
i’d be bullish on anyone doing spot or serverless like beam.cloud
and bearish on anyone buying power and GPUs and calling themselves a neocloud
Andy Long retweeted
Congratulations to @IlhanScheer and @Aleph__Alpha on Kolibri.
78B params, 3.46B active, up to 1M tokens of context, open weights.
Trained by Aleph Alpha on Verda in Europe.
We ran the clusters, built the k8s stack, and contributed to their training efficiency work: verda.com/blog/aleph-alpha-k…
Andy Long retweeted
We’ve been working with the @vllm_project team to accelerate open-source AI.
Thanks to the new DSpark trained on Verda’s GB300 NVL72, Kimi K3 just got ~4x faster in vLLM: ~110 to ~435 tok/s per user on math reasoning.
Full technical report: vllm.ai/blog/2026-09-15-kimi…
Kimi K3 is a 2.8T-parameter model. We trained a DSpark speculator for it, and the speedup holds up.
Single-stream math reasoning goes from ~110 to ~435 tokens/sec per user. Under concurrent load, up to ~3.5x higher output throughput at matched interactivity. The drafter is a 5B model proposing 8 tokens a step, and on math it gets ~6.4 of them accepted per round.
Training a drafter for a model this big meant going multi-node: Speculators plus a new Mooncake connector streaming hidden states between vLLM inference and training over RDMA. Two GB300 nodes to serve the target, one to train.
vllm.ai/blog/2026-09-15-kimi…
Andy Long retweeted
Cloud startup Verda raises $189 million in fresh financing, underscoring continued investor appetite for the infrastructure powering AI bloomberg.com/news/articles/…
Andy Long retweeted
Ever wonder something stupid like whether Kimi K3 does better if you give it more active experts? How does work distribute across different experts - ie. are few doing most of the work?
matrix-multiply.com/posts/ki…
TLDR;
- No, more experts is not automatically better, assuming this is tied to how the model is trained. Maybe there is some easy gains if compute is cheap enough but at least not by doing it the most naive way.
- Some workloads have higher affinity to certain experts, but really depends on what benchmark you're checking.
- You can't run a benchmark once to make conclusions, which can make testing on large models quite compute intense.
Andy Long retweeted
Day 2️⃣ and it’s our turn on the @vllm_project track.
2 PM, Breakout 3: our MLOps Engineer, Ananth will present lessons from serving GLM-5.2 (@Zai_org ) & @Kimi_Moonshot K3 on GB300 & B300 clusters.
Catch the talk, then come say hi at our booth and grab some Finnish chocolate 🍫
Andy Long retweeted
Love seeing this 😍
The @RedHat_AI team built their DSpark speculator (~4x faster Kimi-K3 decoding) on our GB300 and the results speak for themselves.
Speculative decoding is one of the optimization dimensions we actively work on in the AI Lab, so this is exactly the kind of work we love hosting 👇
~4x faster Kimi-K3 decoding.
Our new DSpark speculator takes single-stream interactivity from ~110 to ~435 tok/s/user on math reasoning, delivers ~3.5x the output throughput at matched interactivity under load, and thanks to sliding window attention (2048-token window across all 5 draft layers), our model holds steady acceptance out to 20K context across 10 LongBench domains.
Check it out! huggingface.co/RedHatAI/Kimi…
More details in reply:
Andy Long retweeted
Umberto Eco, who owned 50,000 books, had this to say about home libraries: "It is foolish to think that you have to read all the books you buy, as it is foolish to criticize those who buy more books than they will ever be able to read. It would be like saying that you should use all the cutlery or glasses or screwdrivers or drill bits you bought before buying new ones.
There are some things in life that we need to always have in plenty, even if we will only use a small portion. If, for example, we consider books as medicine, we understand that it is good to have many at home rather than a few: when you want to feel better, then you go to the 'medicine closet' and choose a book. Not a random one, but the right book for that moment. That's why you should always have a nutrition choice!
Those who buy only one book, read only that one and then get rid of it - they simply apply the consumer mentality to books- that is, they consider them a consumer product, a good. Those who love books know that a book is anything but a commodity."
Andy Long retweeted
Everyone who thought the killbots would be human-shaped is going to be be extremely disappointed in their predictive abilities for the approximately .03 seconds they have remaining before the metallic ball of living yarn rams into them at 460mph, boring a hole through their torso
Andy Long retweeted
Very cool, entertaining and with high production values: if I were @Nigel_Farage, I’d consider dipping into that £5 million to try to get out a video that can compare to it.
Andy Long retweeted
Gary Economics has just ruined its credibility and proven that it clearly knows very little about economics.
The absolute tax legend @DanNeidle destroyed Gary in his own documentary.
The very documentary just a compilation of Gary being proven wrong time and time again about his "solution" to economic inequality.
Let me know if you have watched it.
The Dan Neidle debate has gotta be one of the best moments in the show!!
#garystevenson
@Channel4 @garyseconomics