Hugging Face new model alerts in Slack, every weekday

PromptCreate

Every weekday at 8am, check Hugging Face for new model releases from the AI labs on my watchlist and post a short briefing to my #new-models Slack channel.

My watchlist is these Hugging Face organizations: meta-llama, mistralai, Qwen, deepseek-ai, google, and microsoft. For each organization, use the Hugging Face List Models operation filtered by that author, sorted by lastModified in descending order, requesting full metadata. Keep only the repositories created or updated in the last 24 hours, and stop paging through an organization once you reach repos older than that window.

For every repository that passes the time filter, call Get Model to pull its license, pipeline tag, library, parameter size tags, and download count. Use that metadata, together with the model card, to work out what the repo actually is.

Filter out the noise before you write anything. Skip pure quantizations and mirror re-uploads: GGUF, AWQ, GPTQ, MLX and 4-bit or 8-bit repacks of a model that already exists, plus copies of another organization's model. The one exception is when a repo is the first quantization of a genuinely new base model, which is worth keeping. Also skip repos where the only thing that changed was a minor model card or README edit, since those show up as updates but are not news.

Rank whatever survives by significance. A brand new base model ranks above a new fine tune, which ranks above a minor update to an existing model.

Send the briefing to #new-models using the Slack Send a Message operation. Give one line per model in ranked order: the model name, a short phrase describing what it actually is, its license, and a link to huggingface.co followed by the repo id. Use Slack mrkdwn formatting so the message is readable, and bold the model names. Keep each description to a single short sentence, because I want to scan this, not read it.

If nothing meaningful shipped that day, post a single line saying so rather than padding the digest with filler.

Example output

*New open models, Tuesday 4 August* • *deepseek-ai/DeepSeek-V4-Base*. New 236B mixture of experts base model, first release in the V4 line. MIT. huggingface.co/deepseek-ai/DeepSeek-V4-Base • *mistralai/Mistral-Small-3.2-24B-Instruct*. Instruction tuned refresh of Mistral Small, focused on better tool calling. Apache 2.0. huggingface.co/mistralai/Mistral-Small-3.2-24B-Instruct • *Qwen/Qwen3-Coder-7B*. Code specialized fine tune of Qwen3 7B, 32k context. Apache 2.0. huggingface.co/Qwen/Qwen3-Coder-7B Skipped 14 quantized and mirror uploads.

Additional information

What does this prompt do?

Checks a watchlist of major AI labs on Hugging Face every weekday morning for models released or updated in the last 24 hours.
Filters out the noise: repackaged and quantized copies of models that already exist, mirror re-uploads, and minor description edits.
Ranks what is left so brand new models come first and small updates come last.
Posts one Slack message to your team channel with a single line per model: what it is, its license, and a link.

What do I need to use this?

A Hugging Face account. A free one is fine.
A Slack workspace and a channel for the briefing, such as #new-models.
The list of AI labs you want to follow. It starts with Meta, Mistral, Qwen, DeepSeek, Google and Microsoft.

How can I customize it?

Change the watchlist. Add labs like Cohere, NousResearch or AI21, or drop the ones you do not care about.
Change the timing or the destination. Run it every day including weekends, or send the briefing to a different channel or straight to you as a direct message.
Change the filters. Keep every quantized release if you run models locally, or only report models above a certain size or download count.

FAQs

Do I need a paid Hugging Face account?

No. A free account is enough, since everything the briefing reads is public. Signing in rather than browsing anonymously also keeps the daily checks running smoothly.

Which AI labs does it watch?

It starts with Meta, Mistral, Qwen, DeepSeek, Google and Microsoft. You can follow any organization that has a Hugging Face page, so adding or removing labs is just editing a list.

Will it flood the channel every time someone uploads a file?

No, and that is the main point of it. Popular models get copied and repackaged dozens of times in the days after launch, and those copies are suppressed. You only hear about a repackaged version when it is the first one for a genuinely new model.

What happens on a quiet day when nothing ships?

You get a single line saying nothing significant shipped. The briefing never pads itself out with filler to look busy.

Can it post to a private channel or send me a direct message instead?

Yes. It can post anywhere in your workspace, as long as the Slack app has been added to that channel first.

Related templates

Chase past-due JobNimbus invoices with escalating emails

Every weekday, find every unpaid JobNimbus invoice, email each customer one reminder that gets firmer as it ages, and post a receivables summary to Slack.

Agentic Task

Flag at-risk students in Canvas before they fall behind

Every weekday at 7am, your active courses are checked for missing work, silent logins, and slipping grades, with a ranked list sent to your advising channel.