clem &#x1f917;(@ClementDelangue)

Why? - Risk comes from capability asymmetry between attackers and defenders. Open source is what ev...

clem 🤗 (@ClementDelangue) — Mon, 20 Apr 2026 15:11:08 GMT

Why?

- Risk comes from capability asymmetry between attackers and defenders. Open source is what every frontier lab already trains on so defenders get the same AI firepower as attackers. With proprietary code, you're on your own. The biggest risk is someone training a model on your obscure stack and attacking you when no public model exists to defend it.

- AI can now read stripped binaries, so proprietary obscurity barely protects anyone anymore. Most legacy firmware and embedded code is closed, binary-only, and no longer maintained. A huge attack surface that just became legible to AI.

- In a Mythos world, software security becomes a speed race: detection, verification, coordination, patch propagation. Closed-source systems are weaker at all four because they centralize knowledge and action inside a vendor, while open ecosystems distribute both.

- In open source, the defender crowd is usually bigger than the attacker crowd. In closed source, it's the opposite. AI force-multiplication will amplify that imbalance.

And when closed-source systems fail, the blast radius tends to be much larger. They sit behind centralized user, customer, and cloud data, while open source more often runs locally with less data concentration.

Let's go open-source!

clem 🤗@ClementDelangue

In a mythos world (which we are already in), closed-source projects will be 10x more at risk than open-source projects!

💬15🔄11❤️80👀10273📊22

In a mythos world (which we are already in), closed-source projects will be 10x more at risk than op...

clem 🤗 (@ClementDelangue) — Mon, 20 Apr 2026 13:25:59 GMT

In a mythos world (which we are already in), closed-source projects will be 10x more at risk than open-source projects!

💬35🔄24❤️361👀39757📊69

Hugging Face is becoming the platform for agents to use and build AI. Now they can call 1M HF spaces...

clem 🤗 (@ClementDelangue) — Sat, 18 Apr 2026 23:07:35 GMT

Hugging Face is becoming the platform for agents to use and build AI. Now they can call 1M HF spaces to do everything the latest specialized models can do!

Your browser does not support the video tag.

💬34🔄109❤️869👀68194📊215

Looking for some beta testers for our new robotics dataset hosting features. Who's currently hosting...

clem 🤗 (@ClementDelangue) — Fri, 17 Apr 2026 16:55:35 GMT

Looking for some beta testers for our new robotics dataset hosting features. Who's currently hosting large robotics datasets and would be down to try? 🤖🤖🤖

💬32🔄11❤️101👀14346📊42

LeRobot is the unsung hero of open robotics!

clem 🤗 (@ClementDelangue) — Thu, 16 Apr 2026 20:01:43 GMT

LeRobot is the unsung hero of open robotics!

💬1🔄12❤️123👀6318📊17

this is the dataset btw: https://t.co/5Dujbti2Mv

clem 🤗 (@ClementDelangue) — Thu, 16 Apr 2026 18:48:10 GMT

this is the dataset btw: huggingface.co/datasets/badlo…

💬0🔄5❤️7👀2649📊2

Adithya is putting his work where his heart is. If you believe in open research and open AI, join a...

clem 🤗 (@ClementDelangue) — Thu, 16 Apr 2026 18:28:43 GMT

Adithya is putting his work where his heart is.

If you believe in open research and open AI, join a company that actually lives those values, not a closed-source, revenue-maximizing one!

Adithya S K@adithya_s_k

Quick career update:
joined @HuggingFace as a Research Engineer to make RL go brrrrr 🚀

P.S. I’ve been a huge fan of hf ever since I started working on ML and contributing to open source. I can’t imagine what open-source AI would look like without HF. I’ve looked up to so many people here, and it’s truly a pleasure to now work with them.

💬15🔄3❤️389👀40581📊48

You can now visualize Pi traces that you upload on @huggingface! Let's make sharing agent traces 1...

clem 🤗 (@ClementDelangue) — Thu, 16 Apr 2026 17:43:48 GMT

You can now visualize Pi traces that you upload on
@huggingface! 

Let's make sharing agent traces 10x more common to make agent AI more open and collaborative! Also, because it's fun to analyze @badlogicgames's traces 😂😂😂

💬5🔄15❤️127👀26061📊25

Completely agree with Jensen! Not exporting AI out of fear is classic case where the cure would be ...

clem 🤗 (@ClementDelangue) — Thu, 16 Apr 2026 16:50:01 GMT

Completely agree with Jensen!

Not exporting AI out of fear is classic case where the cure would be 100x worse than the disease. You slow down innovation, progress and US technology and economic leadership in the foolish hope that you'll prevent an edge case that hasn't even proved to be dangerous. 

It’s like stopping exports of metal because it might help adversaries build teleportation machines.

Dwarkesh Patel@dwarkesh_sp

The Jensen Huang episode.

00:00 – Is Nvidia’s biggest moat its grip on scarce supply chains?
16:25 – Will TPUs break Nvidia’s hold on AI compute?
41:06 – Why doesn’t Nvidia become a hyperscaler?
57:36 – Should we be selling AI chips to China?
35:06 – Why doesn’t Nvidia make multiple different chip architectures?

Look up Dwarkesh Podcast on YouTube, Apple Podcasts, Spotify, etc. Enjoy!

Your browser does not support the video tag.

💬33🔄15❤️196👀19151📊43

clem 🤗 (@ClementDelangue) — Wed, 15 Apr 2026 17:16:06 GMT

Saint Nomad@Saint_n0mad

which one of them are you most excited for

💬6🔄4❤️135👀26414📊15

Open-source is the solution to cyber-security because with new AI capabilities, all of the open-sour...

clem 🤗 (@ClementDelangue) — Wed, 15 Apr 2026 16:35:09 GMT

Open-source is the solution to cyber-security because with new AI capabilities, all of the open-source repos will be inspected and patched 100x faster/better than any closed-source system!

Let's take a practical example:

You're a startup building a new feature that deals with private user information so you want to be careful. 

Do you trust more:
- one of your team members with 0 security background to build the feature from 100% new AI slope code with their agent in a complete opaque/unauditable way (for example sending data to APIs)
or
- one of your team members with their agent using mature open-source projects where thousands of people and agents have been inspecting the code and vulnerabilities are regularly patched and where the data can stay locally on the user or the company devices

The answer is obvious!

clem 🤗@ClementDelangue

Weird how some people always target open-source in AI!
 
First it was:
“Open-source AI will destroy the world” (spoiler: it didn't and it won't)
Now:
“Open-source is a cybersecurity threat because of AI”

Both narratives are far too simplistic.

The truth is that the exact same risks exist in closed-source systems, often even more so. For example, in practice, APIs can create much bigger data and security vulnerabilities than open systems you can inspect, self-host, and secure yourself.

And as with software more broadly, open-source often ends up more secure because it benefits from far more scrutiny than private internal systems.

The reality is not “open vs closed.”

The reality is that AI is raising cybersecurity stakes across the board, and we need to tackle that seriously together.

💬19🔄20❤️149👀12602📊29

Weird how some people always target open-source in AI! First it was: “Open-source AI will destroy ...

clem 🤗 (@ClementDelangue) — Wed, 15 Apr 2026 16:14:18 GMT

Weird how some people always target open-source in AI!
 
First it was:
“Open-source AI will destroy the world” (spoiler: it didn't and it won't)
Now:
“Open-source is a cybersecurity threat because of AI”

Both narratives are far too simplistic.

The truth is that the exact same risks exist in closed-source systems, often even more so. For example, in practice, APIs can create much bigger data and security vulnerabilities than open systems you can inspect, self-host, and secure yourself.

And as with software more broadly, open-source often ends up more secure because it benefits from far more scrutiny than private internal systems.

The reality is not “open vs closed.”

The reality is that AI is raising cybersecurity stakes across the board, and we need to tackle that seriously together.

💬56🔄70❤️356👀53677📊96

More people should work on harnesses for open and local models!

clem 🤗 (@ClementDelangue) — Wed, 15 Apr 2026 13:20:52 GMT

More people should work on harnesses for open and local models!

eric zakariasson@ericzakariasson

i still think about how these models performed better in cursor harness than their native ones

💬16🔄14❤️178👀29219📊30

let's go open-source and local models!

clem 🤗 (@ClementDelangue) — Tue, 14 Apr 2026 20:07:00 GMT

let's go open-source and local models!

Anissa Gardizy@anissagardizy8

Uber's CTO told @LauraBratton5 that AI coding tools—particularly Anthropic’s Claude Code—has already maxed out its 2026 AI budget 📈

“I'm back to the drawing board, because the budget I thought I would need is blown away already,” Neppalli Naga said.
theinformation.com/newsletters/ap…V

💬12🔄28❤️234👀27421📊32

Is there somewhere a collection of the best agent/coding harnesses for each models, especially open-...

clem 🤗 (@ClementDelangue) — Tue, 14 Apr 2026 19:43:44 GMT

Is there somewhere a collection of the best agent/coding harnesses for each models, especially open-source and local ones?

In my opinion, the biggest reason why people are struggling with open/local models these days is that the agent/coding harnesses in most open agent are not designed for them and expect it to magically work when they switch models from the default.

💬49🔄21❤️266👀29920📊93

just getting started

clem 🤗 (@ClementDelangue) — Tue, 14 Apr 2026 15:32:30 GMT

just getting started

Jannik Malte Meissner 🇺🇦@jannikmeissner

This is great stuff! @huggingface is on a huge roll these days, congrats @ClementDelangue, your team is shipping like they're on fire! 🔥🤗

💬0🔄1❤️14👀4592📊1

Time to follow https://t.co/K4Rf9nWzU8!

clem 🤗 (@ClementDelangue) — Tue, 14 Apr 2026 14:22:03 GMT

Time to follow hf.co/tencent!

Tengfei Wang@DylanTFWang

Genie3 generates videos. We generate 𝟯𝗗 𝘄𝗼𝗿𝗹𝗱𝘀 you can actually use.

Launching tomorrow — #HYWorld#HYWorld 2.0, an engine-ready World Model🚀

This isn't a video. It's a real 3D scene, all generated &  editable. One image in. A whole 3D world out.
🔥Open-source tomorrow

Your browser does not support the video tag.

💬2🔄15❤️222👀44583📊30

Introducing Kernels on the Hugging Face Hub ✨ What if shipping a GPU kernel was as easy as pushing ...

clem 🤗 (@ClementDelangue) — Tue, 14 Apr 2026 14:02:04 GMT

Introducing Kernels on the Hugging Face Hub ✨

What if shipping a GPU kernel was as easy as pushing a model?

- Pre-compiled for your exact GPU, PyTorch & OS
- Multiple kernel versions coexist in one process
- torch.compile compatible
- 1.7x–2.5x speedups over PyTorch baselines

Your browser does not support the video tag.

💬68🔄220❤️1638👀195582📊372

we used Chandra-OCR-2 by @datalabto: https://t.co/vAuYw1JgRd Full write-up by @NielsRogge: https://...

clem 🤗 (@ClementDelangue) — Mon, 13 Apr 2026 20:10:23 GMT

we used Chandra-OCR-2 by @datalabto: huggingface.co/datalab-to/cha…

Full write-up by @NielsRogge:
huggingface.co/blog/nielsr/oc…

💬7🔄17❤️190👀10318📊59

We just OCR'd 27,000 arxiv papers into Markdown using an open 5B model, 16 parallel HF Jobs on L40S ...

clem 🤗 (@ClementDelangue) — Mon, 13 Apr 2026 19:52:46 GMT

We just OCR'd 27,000 arxiv papers into Markdown using an open 5B model, 16 parallel HF Jobs on L40S GPUs, and a mounted bucket.

Total cost: $850 Total time: ~29 hours Jobs that crashed: 0

This now powers "Chat with your paper" on hf.co/papers

💬88🔄247❤️2309👀164991📊586

Should we start an open Glasswing?

clem 🤗 (@ClementDelangue) — Fri, 10 Apr 2026 19:16:24 GMT

Should we start an open Glasswing?

💬126🔄66❤️1183👀91674📊227

In a world where writing code to build websites and apps is trivial (thank you Lovable, Cursor, Clau...

clem 🤗 (@ClementDelangue) — Fri, 10 Apr 2026 15:23:54 GMT

In a world where writing code to build websites and apps is trivial (thank you Lovable, Cursor, Claude,...), the real differentiation for you and your company (and what makes you successful) will be how you manage to train, run and optimize AI models yourself.

That's why at Hugging Face, we're doubling down on enabling more to become AI builders rather than AI users. We're releasing this week Kernels on the Hugging Face hub. This repo type is for the hardcore AI engineers among you.

Kernels are collections of optimized binary operations where hardware providers support is a first-class citizen:
- CUDA
- ROCm
- Apple Silicon
- Intel XPU

Expect to see more of this repo type on Hugging Face in the coming days.

Featured here: the Flash Attention kernel from @sgl_project team ❤️

💬28🔄48❤️372👀55700📊104

Feels like one of the cybersecurity risks over the coming months will be widely used open-source pro...

clem 🤗 (@ClementDelangue) — Thu, 09 Apr 2026 13:46:57 GMT

Feels like one of the cybersecurity risks over the coming months will be widely used open-source projects that are simply too lightly maintained for how critical they’ve become.

A few ways to help:
- fund open source more, and reward maintainers better
- bring some of the most critical projects under more resourced umbrellas like @linuxfoundation or @huggingface

Which projects would actually make sense to join HF if we wanted to help them on cybersecurity?

💬30🔄12❤️142👀17704📊43

@sundarpichai @abidlabs #1 trending on @huggingface (+ many variants in the top 10). Fantastic contr...

clem 🤗 (@ClementDelangue) — Thu, 09 Apr 2026 00:35:13 GMT

@sundarpichai@abidlabs#1 trending on @huggingface (+ many variants in the top 10). Fantastic contribution!

💬2🔄4❤️114👀10205📊13

"But here is what we found when we tested: We took the specific vulnerabilities Anthropic showcases ...

clem 🤗 (@ClementDelangue) — Wed, 08 Apr 2026 18:58:08 GMT

"But here is what we found when we tested: We took the specific vulnerabilities Anthropic showcases in their announcement, isolated the relevant code, and ran them through small, cheap, open-weights models. Those models recovered much of the same analysis. Eight out of eight models detected Mythos's flagship FreeBSD exploit, including one with only 3.6 billion active parameters costing $0.11 per million tokens. A 5.1B-active open model recovered the core chain of the 27-year-old OpenBSD bug." aisle.com/blog/ai-cybers…

💬110🔄344❤️2443👀721053📊563

"The priority for defenders is to start building now: the scaffolds, the pipelines, the maintainer r...

clem 🤗 (@ClementDelangue) — Wed, 08 Apr 2026 18:55:02 GMT

"The priority for defenders is to start building now: the scaffolds, the pipelines, the maintainer relationships, the integration into development workflows. The models are ready. The question is whether the rest of the ecosystem is."

aisle.com/blog/ai-cybers…

💬12🔄11❤️42👀4943📊17

@steipete 100%! the open-source builders and maintainers should be celebrated and rewarded 1000x mor...

clem 🤗 (@ClementDelangue) — Wed, 08 Apr 2026 16:11:32 GMT

@steipete 100%! the open-source builders and maintainers should be celebrated and rewarded 1000x more!

💬2🔄6❤️119👀16026📊14

Cool thing is that you might be safer in jail with local AI than anywhere else!

clem 🤗 (@ClementDelangue) — Wed, 08 Apr 2026 14:50:14 GMT

Cool thing is that you might be safer in jail with local AI than anywhere else!

Elizabeth Holmes@ElizabethHolmes

Delete your search history, delete your bookmarks, delete your reddit, medical records, 12 yr old tumblr, delete everything.

Every photo on the cloud, every message on every platform.

None of it is safe. 

It will all become public in the next year
Local storage and compute 📈

💬7🔄2❤️88👀18294📊12

Glad to see that Mytos is putting back cybersecurity in the spotlight. Our main contribution to the ...

clem 🤗 (@ClementDelangue) — Wed, 08 Apr 2026 14:33:13 GMT

Glad to see that Mytos is putting back cybersecurity in the spotlight. Our main contribution to the topic has been the creation of Safetensors by @narsilou@huggingface in collaboration with @AiEleuther@StabilityAI.

It has become the most popular way to share models safely and today @LysandreJik announced at #PyTorchCon that the project is joining the @PyTorch foundation to continue to scale (for example with torch core integration).

Let's turn talks about cyber-security into action!

💬6🔄8❤️71👀8317📊14

Anthropic had the most powerful cyber-security model in the history of this world and their internal...

clem 🤗 (@ClementDelangue) — Wed, 08 Apr 2026 14:24:50 GMT

Anthropic had the most powerful cyber-security model in the history of this world and their internal code based still leaked? 

We should assume everyone can be compromised, and build systems that keep the cost of attacking higher than the reward, limit blast radius when attacks succeed, create fast repair loops after weaknesses are found and reduce systemic risk.

Open-source will play a major role in all of that!

💬80🔄68❤️666👀64968📊152

Very cool open-source traces from @TheZachMueller @LambdaAPI: https://t.co/N1MbsQsXLW 150M tokens f...

clem 🤗 (@ClementDelangue) — Tue, 07 Apr 2026 17:57:18 GMT

Very cool open-source traces from @TheZachMueller@LambdaAPI: huggingface.co/datasets/lambd…

150M tokens for @NousResearch's Hermes harness with Kimi-K2.5 & GLM 5.1 that was just released!

💬12🔄35❤️260👀50598📊60

The best performing model on SWE-Bench Pro is open-source on @huggingface! Welcome GLM 5.1! https:/...

clem 🤗 (@ClementDelangue) — Tue, 07 Apr 2026 16:31:38 GMT

The best performing model on SWE-Bench Pro is open-source on @huggingface! Welcome GLM 5.1!

huggingface.co/zai-org/GLM-5.1

💬30🔄75❤️634👀36164📊113

Curious, how much are you all spending in S3/R2 or storage in general these days?

clem 🤗 (@ClementDelangue) — Tue, 07 Apr 2026 14:51:46 GMT

Curious, how much are you all spending in S3/R2 or storage in general these days?

💬13🔄0❤️41👀12608📊15

@pmarca Let’s go open-source and local AI!

clem 🤗 (@ClementDelangue) — Tue, 07 Apr 2026 13:50:57 GMT

@pmarca Let’s go open-source and local AI!

💬5🔄8❤️150👀18542📊20

@theo can you help?

clem 🤗 (@ClementDelangue) — Mon, 06 Apr 2026 16:24:35 GMT

@theo can you help?

💬2🔄2❤️37👀4882📊5

We keep saying we want open-source frontier agents. Fine. Then let’s build the dataset. @badlogicg...

clem 🤗 (@ClementDelangue) — Mon, 06 Apr 2026 16:22:43 GMT

We keep saying we want open-source frontier agents.

Fine. Then let’s build the dataset.

@badlogicgames, creator of Pi, just shared some of his agent traces used to build Pi on @huggingface. I’m now sharing some of mine too, exporting them from @hermes, @opencode, and Claude via @tracesdotcom, and I’ll keep going.

Why this matters: one of the biggest bottlenecks for open-source agent models is the data. And all of us are generating that data every day through our conversations with agents.

If enough builders share even a slice of their traces publicly, we can create the largest crowdsourced open dataset for agents.

Time to put your tokens where your mouth is and give a chance for open source to win!

💬49🔄133❤️821👀99133📊215

We keep saying we want open-source frontier agents. Fine. Then let’s build the dataset. @badlogicg...

clem 🤗 (@ClementDelangue) — Mon, 06 Apr 2026 15:28:03 GMT

We keep saying we want open-source frontier agents.

Fine. Then let’s build the dataset.

@badlogicgames, creator of Pi, just shared some of his agent traces used to build Pi on @huggingface. I’m now sharing some of mine too, exporting them from @hermes, @opencode, and Claude via @tracesdotcom, and I’ll keep going.

Why this matters: one of the biggest bottlenecks for open-source agent models is the data. And all of us are generating that data every day through our conversations with agents.

If enough builders share even a slice of their traces publicly, we can create the largest crowdsourced open dataset for agents.

Time to put your tokens where your mouth is, create the biggest crowd-sourced agent dataset and give a chance for open source to win!

💬14🔄25❤️143👀8434📊44

Gemma 4 is #1 on @huggingface!

clem 🤗 (@ClementDelangue) — Sun, 05 Apr 2026 21:55:06 GMT

Gemma 4 is #1 on @huggingface!

💬35🔄67❤️856👀59466📊124

@garrytan Let’s go open-source and local!

clem 🤗 (@ClementDelangue) — Sat, 04 Apr 2026 15:26:20 GMT

@garrytan Let’s go open-source and local!

💬17🔄10❤️138👀24332📊29

I think it’s @NaveenGRao who said it before but wouldn’t be surprised if the frontier labs cut their...

clem 🤗 (@ClementDelangue) — Sat, 04 Apr 2026 14:36:33 GMT

I think it’s @NaveenGRao who said it before but wouldn’t be surprised if the frontier labs cut their APIs entirely at some point. In a compute constrained world, they’ll always prioritize their own direct products/customers. Makes it scary and unsustainable to only build on top of their APIs!

💬55🔄36❤️496👀72607📊114

Let’s go @arcee_ai (and startups in general)!

clem 🤗 (@ClementDelangue) — Sat, 04 Apr 2026 14:31:40 GMT

Let’s go @arcee_ai (and startups in general)!

💬8🔄8❤️144👀10215📊25

llama-server -hf ggml-org/gemma-4-26b-a4b-it-GGUF:Q4_K_M openclaw onboard --non-interactive \ --a...

clem 🤗 (@ClementDelangue) — Sat, 04 Apr 2026 00:18:18 GMT

llama-server -hf ggml-org/gemma-4-26b-a4b-it-GGUF:Q4_K_M

openclaw onboard --non-interactive \
  --auth-choice custom-api-key \
  --custom-base-url "http://127.0.0.1:8080/v1" \
  --custom-model-id "ggml-org-gemma-4-26b-a4b-gguf" \
  --custom-api-key "llama.cpp" \
  --secret-input-mode plaintext \
  --custom-compatibility openai \
  --accept-risk

Peter Steinberger 🦞@steipete

woke up and my mentions are full of these

Both me and @davemorin tried to talk sense into Anthropic, best we managed was delaying this for a week.

Funny how timings match up, first they copy some popular features into their closed harness, then they lock out open source.

💬26🔄50❤️560👀116570📊121

Time to move to open or local models from Hugging Face! All instructions are here: https://t.co/0w0...

clem 🤗 (@ClementDelangue) — Sat, 04 Apr 2026 00:06:54 GMT

Time to move to open or local models from Hugging Face! All instructions are here:  huggingface.co/blog/liberate-…

Boris Cherny@bcherny

Starting tomorrow at 12pm PT, Claude subscriptions will no longer cover usage on third-party tools like OpenClaw.

You can still use these tools with your Claude login via extra usage bundles (now available at a discount), or with a Claude API key.

💬25🔄72❤️432👀70360📊88

First public PR in years from the legend @JeffDean was to @huggingface transformers. Any other proud...

clem 🤗 (@ClementDelangue) — Fri, 03 Apr 2026 22:53:38 GMT

First public PR in years from the legend @JeffDean was to @huggingface transformers. Any other proud moment for us and our community!

Omar Sanseviero@osanseviero

💬7🔄22❤️392👀48803📊39

If it’s not either pushing the frontier meaningfully or open-source, no one will care these days (wh...

clem 🤗 (@ClementDelangue) — Fri, 03 Apr 2026 09:40:36 GMT

If it’s not either pushing the frontier meaningfully or open-source, no one will care these days (which is why most orgs should release open-source to get some attention and developer mindshare)

Harveen Singh Chadha@HarveenChadha

The AI race is so brutal that Microsoft released 3 new models yesterday and nobody noticed

💬22🔄13❤️218👀17542📊33

Google just re-entered the game 🔥🔥 They want to take the crown 👑 back from Chinese open source A...

clem 🤗 (@ClementDelangue) — Fri, 03 Apr 2026 05:41:00 GMT

Google just re-entered the game 🔥🔥

They want to take the crown 👑 back from Chinese open source AI.

And... Gemma 4 is FINALLY Apache 2.0 aka real-open-source-licensed.

From what I've seen it's going to be a pretty significant model. But give it a try yourself today:

brew install llama.cpp --HEAD

if you have at least 24GB of RAM or VRAM, run the (very good) 26B MOE:

llama-server -hf ggml-org/gemma-4-26B-A4B-it-GGUF:Q4_K_M

if you have 16GB of RAM or VRAM, run the dense E4B:

llama-server -hf ggml-org/gemma-4-E4B-it-GGUF:Q8_0

💬34🔄64❤️588👀36873📊136

Will @tbpn cover the @google gemma 4 release today?

clem 🤗 (@ClementDelangue) — Thu, 02 Apr 2026 20:32:22 GMT

Will @tbpn cover the @google gemma 4 release today?

💬12🔄3❤️266👀24193📊34

this is Gemma 4 running locally on a 3 year old mac meaning: - free (=$0 no matter how much you use)...

clem 🤗 (@ClementDelangue) — Thu, 02 Apr 2026 19:48:01 GMT

this is Gemma 4 running locally on a 3 year old mac meaning:
- free (=$0 no matter how much you use)
- safe (you're not leaking all your data via unsafe APIs)
- fast (as you can see)

Georgi Gerganov@ggerganov

Let me demonstrate the true power of llama.cpp:

- Running on Mac Studio M2 Ultra (3 years old)
- Gemma 4 26B A4B Q8_0 (full quality)
- Built-in WebUI (ships with llama.cpp)
- MCP support out of the box (web-search, HF, github, etc.)
- Prompt speculative decoding

The result: 300t/s

(realtime video)

Your browser does not support the video tag.

💬36🔄50❤️745👀138726📊152