vast.ai

Vast is the infrastructure layer where AI agents autonomously design, procure, and optimize their own compute.

[vast.ai @vast_ai]

https://t.co/2WHvjspuyh is hiring.

We're growing. Fast.

Agent-ready AI infrastructure. Profitable, venture-backed. New roles across engineering, research + DevRel, mostly deep systems + GPU work. On-site in SF or LA.

https://t.co/mexpPT9oZq

[vast.ai @vast_ai]

Book with us here: https://t.co/TRShsU09cd

[vast.ai @vast_ai]

https://t.co/DfwOtzr83c is at The AI Conference 2026.

September 29 to October 1. Pier 48, San Francisco. Booth 202.

If you are planning GPU capacity for the next two quarters, book time with us. Bring the workload and the timeline. You leave with real pricing and availability across our network.

Booking link below.
@AIconference

[vast.ai @vast_ai]

Renting GPU compute beats buying when your LLM workload is bursty or changing.

No hardware spend. Pick the GPU for the job. Pay only while it's running.

Local hosting only wins if your workload is steady and predictable.

https://t.co/wOjVOJsvSl

[Travis Cannell @traviscannell]

@vast_ai is over 90% utilized! Amazing seeing the fleet cooking.

Bring your GPUs asap to realize the best on-demand pricing.

[vast.ai @vast_ai]

Nvidia reportedly agreed to buy Hugging Face for $12.9B. Last funding round: $4.5B valuation. They turned down Nvidia's own $500M offer at $7B last year.

The gap says a lot about how much owning open-model distribution is suddenly worth.

More here: https://t.co/VF3PXGoaVF

[vast.ai @vast_ai]

@its_d_i_m_a @Alibaba_Qwen @opencode 🦾

[vast.ai @vast_ai]

@eb70899 @Meta something on your mind? 👀

[vast.ai @vast_ai]

New Muse Glimmer 30B template available in the model library

Muse Glimmer 30B by @Meta is a multimodal reasoning model built for autonomous agentic work. It combines multi-step reasoning, reliable tool use, image and document understanding, and failure recovery in a single dense model, targeting agents that run end-to-end workflows rather than one-shot completions.

Learn more here: https://t.co/eTXslmCbgj

[vast.ai @vast_ai]

MiniMax H3 Now Available in the Model Library

@MiniMax_AI H3 is an omni-modal generative system from MiniMaxAI that produces video with native stereo audio from text, images, video, and audio inputs. It understands multimodal context as a single unified sequence rather than treating each input type separately, and predicts video and audio latents jointly in one forward pass so speech, sound effects, and music land in sync with the picture.

Learn more about the MiniMax H3 template in the model library: https://t.co/2C9BQu3GFg

[vast.ai @vast_ai]

DeepSeek V4 Flash Now Available in the Model Library

DeepSeek V4 Flash is a Mixture-of-Experts (MoE) language model with 284B total parameters and 13B activated per token. This is the official release, which supersedes the earlier preview and substantially enhances agentic capability while keeping the same model structure and size. It targets highly efficient long-context intelligence, supporting a context window of up to one million tokens, and is a strong general-purpose model for reasoning, coding, and agentic workflows.

Learn more about the DeepSeek V4 Flash template in the model library: https://t.co/Vw5h6JeIaZ

[vast.ai @vast_ai]

New on Vast: hundreds of B200s live now with more coming through September, plus Kimi K3 (the first open-weight 3T-class model) and other new templates in the Model Library.

Also shipped: webhooks, per-event notifications, and Hugging Face Storage Bucket support.

Full rundown: https://t.co/yTTulR2Tgh

[Julien Chaumond @julien_c]

Do you use @vast_ai to spawn GPU instances from a large marketplace of available machines?

If you do, you can now mount HF Storage Buckets to your instances and unlock super fast/scalable storage, optimized for AI 🔥🔥

Great collab, give it a try its super convenient 🥰

[vast.ai @vast_ai]

@abidlabs Very exciting!

[vast.ai @vast_ai]

Hugging Face Storage Buckets are now on https://t.co/DfwOtzr83c

Connect your HF Storage Bucket as a Cloud Connection in your Vast settings, and every GPU instance you rent can pull datasets and checkpoints straight from your bucket and push results back. No manual transfers, no re-uploads.

Your data lives on Hugging Face. Your compute runs on Vast GPUs.

Available now for all users.

Docs:
Hugging Face Storage Buckets: https://t.co/Nnrt9noPKh
Vast Cloud Sync: https://t.co/mRKcIem1x5

@huggingface

[vast.ai @vast_ai]

NVIDIA's DGX Spark packs a Grace Blackwell GPU and 128GB unified memory into a box the size of a Mac Mini. Early reviews are mixed though, with reports of thermal throttling cutting into performance. We broke down the specs and what reviewers are actually seeing.

https://t.co/BZ2GH481hF

[vast.ai @vast_ai]

Kimi K3 template now available in the https://t.co/DfwOtzr83c model library

Kimi K3 is an open-weight, native multimodal agentic model from @Kimi_Moonshot their most capable model to date. It is a 2.8T-parameter Mixture-of-Experts model built on Kimi Delta Attention (KDA) and Attention Residuals (AttnRes), with native vision capabilities and a 1-million-token context window. It is the first open 3T-class model, designed for frontier intelligence across long-horizon coding, knowledge work, and reasoning.

Learn more about the Kimi K3 template in the model library: https://t.co/kFu9cnvafz

[vast.ai @vast_ai]

NVIDIA DGX Spark: 6"x6"x2", 128GB unified memory, models up to 200B parameters. A supercomputer that fits in your hand.

Impressive hardware. But compute needs tend to outgrow a fixed box.

On Vast, rent GPUs that exceed Spark's specs, on demand, at 5-6X savings vs traditional cloud.

Full breakdown: https://t.co/BZ2GH481hF

[vast.ai @vast_ai]

@christianelton yeah, hey, since 2017

[vast.ai @vast_ai]

@BearMarketMiner 👀

[vast.ai @vast_ai]

How much does a GPU actually cost to rent right now?

We put together a live pricing guide comparing real-time rates across RTX 5090, H100, H200, B200, B300 — plus how https://t.co/DfwOtzqAdE stacks up against RunPod, Lambda, and CoreWeave's published pricing.

https://t.co/vcphaJLCsU