dstack @dstackai

The orchestration layer for modern ML teams. Run development, training, & inference jobs on GPU - whether on hyperscalers, neoclouds, or your on-prem hardware dstack.ai Joined January 2020

Tweets

827
Followers

1K
Following

29
Likes

651

Andrey Cheptsov @andrey_cheptsov

a day ago

🚀 @dstackai is adding support for NVIDIA B200 spot instances on @googlecloud — testing it now, coming soon! Next up: support for GCP’s Dynamic Workload Scheduler – stay tuned!

0 1 5 241 0

Download Image

knarik @liracinai

a day ago

signups are going wild. but we’re still looking for GPU experts and AI researchers in SF 👀 luma.com/tmpnldk5

0 2 2 247 0

Download Image

How do Prefill–Decode ratios impact GPU inference performance? We benchmarked gpt-oss-120b with @sgl_project on @nvidia (also applies to @AMD). Turns out, a fixed ratio is robust — no need for dynamic adjustment. Read more 👉 dstack.ai/blog/benchmark…

0 3 6 1K 2

dstack @dstackai

4 days ago

We benchmarked Prefill–Decode ratios. Why: dynamic ratio planning adds complexity — do we actually need it? What we found: fixed ratios perform robustly. Why it matters: simpler orchestration, fewer moving parts, same performance. Read: dstack.ai/blog/benchmark…

0 4 14 2K 3

Andrey Cheptsov @andrey_cheptsov

4 days ago

If you’ve tried Prefill–Decode inference on @nvidia or @AMD GPUs (Kubernetes or not), we’d love to hear your take!

dstack @dstackai

4 days ago

If you’ve tried Prefill–Decode inference on @nvidia or @AMD GPUs (Kubernetes or not), we’d love to hear your take!

0 4 14 2K 3

0 3 6 911 0

Hot Aisle @HotAisle

5 days ago

Got some feedback today that our Quick Start page should have some links to the AMD documentation to get started, so I just whipped something up. What do you think? Anything else we should add? hotaisle.xyz/quick-start/

3 2 10 587 1

Andrey Cheptsov @andrey_cheptsov

a week ago

@HotAisle @dstackai Great to see @AMD GPUs making their way to the cloud — accessible on-demand and at a great price point. Hoping @HotAisle joins the @dstackai Sky marketplace next! @AnushElangovan time to spread the word 🚀

0 1 2 132 0

Hot Aisle @HotAisle

a week ago

Hey AI Developers! Our friends at @dstackai now support our cloud natively through our API: admin.hotaisle.app/api/docs/. That means you can automate the spin up and tear down of compute on our VMs instantly, no contracts, no friction, fully on-demand. We’re flipping the GPU compute…

3 4 16 1K 4

dstack @dstackai

2 weeks ago

Big update for the dstack Sky GPU marketplace 🎉 @nebiusai joins as the first provider with on-demand and spot clusters. Until now, Nebius GPUs were accessible only through the open-source control plane. 👉🏼More on how dstack helps you manage clusters and accelerate training &…

1 5 2 689 0

Download Image

knarik @liracinai

2 weeks ago

the best feeling is when users totally understand what we’re building 💜

0 2 3 453 1

Download Image

dstack @dstackai

2 weeks ago

🚀 dstack 0.19.29 is here! This release takes another step toward simpler, fully declarative GPU fleet management. dstack now automatically reuses existing fleets when they fit. We've also improved the Offers UI, which now supports grouping by backend for easier comparison.…

0 2 6 1K 1

Download Image

dstack @dstackai

2 weeks ago

Don’t want to self-host the open-source dstack? Try dstack Sky. Today’s release adds a project wizard to choose between the GPU marketplace or your own cloud accounts. New accounts also get free GPU credits.

1 5 7 2K 1

Download Video

Sergei Egorov 🦋 @bsideup

2 weeks ago

It has never been easier to try @dstackai! I heard that it takes less than 5 minutes, btw 🤓

dstack @dstackai

2 weeks ago

It has never been easier to try @dstackai! I heard that it takes less than 5 minutes, btw 🤓

1 5 7 2K 1

Download Video

1 3 10 2K 2

chansung @algo_diver

2 weeks ago

one of the best "centralized" management of cloud & on-prem GPUs. This is my go-to open source toolkit to run a lot of experiments in parallel w/ cheapest possible GPU instances on the market.

dstack @dstackai

2 weeks ago

one of the best "centralized" management of cloud & on-prem GPUs. This is my go-to open source toolkit to run a lot of experiments in parallel w/ cheapest possible GPU instances on the market.