The exclusive partnership includes hosting private runs and agent traces
Today we're announcing a multi-million-dollar strategic partnership with @huggingface.
The Hub is now the exclusive home for everything we build: every open model we release, every private run, every proprietary dataset, and every agent trace. All of it lives there.
Positive users praise Arcee AI's switch from AWS S3 to Hugging Face Private Storage as a smart move for open-source independence and efficiency, while negative users call the partnership risky dependency on one platform.
More details here: https://huggingface.co/blog/clem/arcee-hf
@ClementDelangue @arcee_ai wow exciting! Do they have opensource models?
Super excited to announce that @arcee_ai is the first major American AI lab to replace AWS S3 with Hugging Face for ALL their models and datasets, public AND private 🔥🔥🔥
Multi-million $ partnership to support American open-source AI, let’s go!
@raulizahi @arcee_ai @code_star @huggingface sure
we are not VC funded, only have funding from angels but gcp/aws/azure will only get you a reasonable amount of credits if u come with an accelerator
@arcee_ai @code_star @huggingface waow care to spare a million for a fella building stuff?
With 14 researchers and about 30 people total across the company - Arcee is a lean operation.
To consistently ship world-class work at our size, infrastructure has to be invisible. The moment our team is spending energy on storage architecture and cost instead of model design, we’ve already lost.
@tekbog @arcee_ai @huggingface Couldn’t help myself.
Read the full partnership announcement on the Arcee blog: https://www.arcee.ai/blog/why-we-made-hugging-face-the-home-for-everything-we-build
@ClementDelangue @arcee_ai two of the best companies championing open-source AI. can't really ask for more. ❤️🫡
@kevincodex @ClementDelangue @arcee_ai Yes, the Trinity family is amazing: https://huggingface.co/arcee-ai/collections
@ClementDelangue @arcee_ai let's go!
Hugging Face Buckets solves this.
It gives us optimized, per-TB storage with egress and CDN included. It makes us completely compute agnostic. We can train wherever cluster capacity is cheapest and best, spin the compute down, and our data is right there waiting.
We punch above our weight because we are ruthless about where we spend our focus.
This partnership removes an entire category of operational overhead. Every ounce of energy we get back goes straight into data quality, training runs, and product experience.
Follow arcee-ai on the Hub and stay tuned, we can’t wait to show you our next generation of open models when they’re ready.
@huggingface @ClementDelangue @arcee_ai oh wow! Thank you HF! Will check it out
This builds on a foundation we've been laying for over two years.
With 200+ models and millions of downloads, Arcee is already one of the most active American labs on the Hub. Making it our official home allows us to meet 15 million ML engineers exactly where they already build.
@raulizahi @arcee_ai @code_star @huggingface the company is http://novlabs.ai but the main product is http://clankercloud.ai
right now we are focusing on being an ai workspace fixing sdlc (k8s supported etc) but are moving into agent orchestrator and an actual cloud
will be doing evals with X and working with Y now
@MaziyarPanahi @arcee_ai
@ClementDelangue @arcee_ai congrats
yea we have people who believe in us as well
the money is for: 1. systems engineering research (we want to have open source underneath like our clanker project and work more on it) but that requires time, compute and credits 2. start dipping our toes into being an actual ai native cloud provider: we are trying to get into startup programs but it’s hard to get answers when u are not coming from a vc firm 3. sys engineering evals: cool stuff but costs too much, we will be working with deepmind soon 4. hire/expand: we dont wanna hire/fire or give people false promises until we are alive by default
well, there are others ofc but that’s just at the top of my head
right now working on agent to agent communication with an orchestrator in shared spaces and waiting on an enterprise contract with a hyperscaler, it’s about research though - eventually to build a world model but that’s way way ahead in the future
cheers if u have a look
i dont see the language performance a bottleneck on local hardware but rather than inference from models that can run it - unless u are referring to this?
the new meta is going to be with local smaller models then delegating into bigger models in sandboxes on the cloud, then communicating both through agents
im pretty sure this is what google and apple have decided together as well. so u have local inference then a small eval to determine the route/intent but working with local models and agents just puts you into the fun space of supporting edge cases for different machines running different software
as for the language question, yea everything can be optimized but as with all engineering is just evaluating if it’s worth spending the time on it, for us, as a company just starting we don’t have that privilege yet
