These agents actively train models and launch Hub Spaces.
why are the open tools so low in the list? We need to improve integration between open platforms and open models @steipete @thdxr @Teknium @badlogicgames!
Coding agents are real users of the @huggingface Hub!
They're searching for models, building and pushing datasets, training models on Jobs, spinning up Spaces...
Now there's public data: each agent's share of Hub traffic, updated monthly 👇
Many users praise open tools and coding agents on Hugging Face for their adaptability, ease of use, and compounding value through better integration, while a few note that closed options still win on workflow convenience.
No Digg Deeper questions have been answered for this story yet.
@ClementDelangue @thdxr @Teknium @badlogicgames Working on that! Building a competitive harness is hard.
Thanks for the help with letting @onusoz join our maintainer team - much appreciated!
why are the open tools so low in the list? We need to improve integration between open platforms and open models @steipete @thdxr @Teknium @badlogicgames!
Coding agents are real users of the @huggingface Hub!
They're searching for models, building and pushing datasets, training models on Jobs, spinning up Spaces...
Now there's public data: each agent's share of Hub traffic, updated monthly 👇
@steipete @thdxr @Teknium @badlogicgames @onusoz thank you! appreciate all your work and collaboration, this is exciting!
@ClementDelangue @thdxr @Teknium @badlogicgames Working on that! Building a competitive harness is hard.
Thanks for the help with letting @onusoz join our maintainer team - much appreciated!
@vanstriendaniel @huggingface what % of total usage is the agent usage? that's the really interesting thing to watch I guess, and to view the trend lines on that?
Explore it in the Dataset Viewer, no code needed:
- the monthly leaderboard + how it shifts as new tools launch - request share vs user share (a few heavy pipelines vs many light users) - daily data: launch spikes, weekday vs weekend patterns...
https://huggingface.co/datasets/huggingface/agent-usage
@strickvl @huggingface Quite hard to count! Some numbers here already https://huggingface.co/blog/hf-cli-for-agents#ai-agent-traffic-on-the-hub
@ClementDelangue @steipete @thdxr @Teknium @badlogicgames Best part about Hermes is it's really easy to /learn new skills. I had it make a Hugging Face CLI skill just for this occasion. Ask your agent to install it!
@vanstriendaniel @huggingface 🧐
@ClementDelangue @steipete @thdxr @Teknium @badlogicgames uh...isn't codex open? #2
Build an agent harness? 37% of agent traffic is still "unknown". Registering a new agent is one small PR: https://huggingface.co/docs/hub/agents-overview#register-your-agent-harness
@ClementDelangue @steipete @thdxr @Teknium @badlogicgames How do I register my harness?
@steipete @ClementDelangue @thdxr @Teknium @badlogicgames @onusoz Open Sauce
@ClementDelangue @steipete @thdxr @Teknium @badlogicgames I'm part of the unknown unregistered
@ClementDelangue @julien_c @steipete @thdxr @Teknium @badlogicgames Seems like this chart could use more breakdown information. Like are these huge users corps, and what percentages of these stats are they pulling. Bigger computers with giant budgets often are what I'm expecting to see here. And that will explain things here better, I believe.
@ClementDelangue @steipete @thdxr @Teknium @badlogicgames guess they track whats popular not whats open. harder to surface when most devs chase top results tbh
@strickvl @huggingface I personally think it will be majority traffic by the end of the year, but I might be way off!
@ClementDelangue @steipete @thdxr @Teknium @badlogicgames As a heavy AI user I think it's a UI + GTM issue. The frontier labs do all things well (strong models) and UI (okay some UI isn't great) and marketing (like they all have our attention for better or worse).
@ClementDelangue @steipete @thdxr @Teknium @badlogicgames Give us more free hours. Lots of technical debt for all.
@ClementDelangue @steipete @thdxr @Teknium @badlogicgames kinda wild that open tools still live under closed ones in practice
the irony writes itself
@vanstriendaniel @_akhaliq @huggingface what do you doing on searching and trying dufferent models? all models technically same, difference on learning data. and tuning for task. nay be is better to search for learning data?
