Pinned Tweet
for daily papers, authors can directly submit them here: huggingface.co/papers/submit
94
48
651
642,969
AK retweeted
Now 10,000 Microduck sold since the Jensen keynote demo.
It’s INSANE!
It’s INSANE!
The Jensen effect 🦆
launch day: about 3,000 MicroDucks sold

24h after @JensenHuang showed it on stage at @nvidia GTC: over 4,000 newly sold units and keep counting
bigger than our (already massive) launch day!
13
14
153
13,623
AK retweeted
Tencent released WeVisOne-VL on Hugging Face
One VLM for unified visual perception, from points and boxes to masks and alpha mattes. It handles detection, counting, OCR, segmentation and matting, and reliably rejects absent targets.
One VLM for unified visual perception, from points and boxes to masks and alpha mattes. It handles detection, counting, OCR, segmentation and matting, and reliably rejects absent targets.
6
9
68
7,018
AK retweeted
Tencent just released Hy-Nomas-Bench on Hugging Face
280 test cases for interaction controllability in video world models, covering multi-turn text, camera navigation, interleaved control, and spatial memory.
280 test cases for interaction controllability in video world models, covering multi-turn text, camera navigation, interleaved control, and spatial memory.
5
6
23
5,091
AK retweeted
We started Mubit in Aberdeen. No network, no lab, nobody knew us. We just wanted AI systems that actually work.
Today we work with some of the best companies and engineers in the world.
And the thing we kept seeing: you and your competitor run the same AI. It should be yours. the next 48 hours for $100 free credits. Enterprises get the first month free.
Today we work with some of the best companies and engineers in the world.
And the thing we kept seeing: you and your competitor run the same AI. It should be yours. the next 48 hours for $100 free credits. Enterprises get the first month free.
200
193
1,101
2,638,948
AK retweeted
MOSS image FAST AS HELL TURBO by @_akhaliq
50 steps → 4 steps ⚡
turbo feels like a sweet spot for generating lots of images fast
regular text posters still look pretty good too. tiny lettering gets a bit blurry, so i'd stick with the original SFT for text-heavy designs!
steps 🆚 4-step version side by side comparison👇
50 steps → 4 steps ⚡
turbo feels like a sweet spot for generating lots of images fast
regular text posters still look pretty good too. tiny lettering gets a bit blurry, so i'd stick with the original SFT for text-heavy designs!
steps 🆚 4-step version side by side comparison👇
4
7
47
6,682
AK retweeted
You know us as your creative asset toolkit.
Now meet @_MonoDesk, a whole new app from our team, built to handle the boring bits of creative work.
Less time on proposals, chasing approvals and sending invoices. More time on the creative side of your business.
Now meet @_MonoDesk, a whole new app from our team, built to handle the boring bits of creative work.
Less time on proposals, chasing approvals and sending invoices. More time on the creative side of your business.
91
97
1,454
1,750,940
AK retweeted
AKV: Cache Engineering for Agents
A new serving system that lets LLM agents directly manage their KV cache — dropping unused history and repositioning retained keys without rewriting or re-prefilling. Reduces average cost by 27.7% while improving task accuracy by 1.7 points.
A new serving system that lets LLM agents directly manage their KV cache — dropping unused history and repositioning retained keys without rewriting or re-prefilling. Reduces average cost by 27.7% while improving task accuracy by 1.7 points.
7
6
75
7,908
AK retweeted
Introducing Canva Craft 🚀
Extraordinary image generation is now available to everyone, powered by our first frontier model built specifically for design.
With Craft, Canva is now the world's second most advanced AI lab for image generation on @DesignArena's global leaderboard.
See what it can do 👇
Extraordinary image generation is now available to everyone, powered by our first frontier model built specifically for design.
With Craft, Canva is now the world's second most advanced AI lab for image generation on @DesignArena's global leaderboard.
See what it can do 👇
251
419
3,283
1,898,030
AK retweeted
Comparing multiple models has never been this easy, thanks to gr.Workflow👇
We used ML Intern to distill FLUX.3 Image Klein's 9B prompt upsampler into 0.8B and 2B models. Short prompt in → full prompt + bounding boxes out. The 0.8B can even run on a CPU.
Links in reply.
We used ML Intern to distill FLUX.3 Image Klein's 9B prompt upsampler into 0.8B and 2B models. Short prompt in → full prompt + bounding boxes out. The 0.8B can even run on a CPU.
Links in reply.
2
7
9
4,799
AK retweeted
So proud to share that @huggingface is giving OpenEnv to the AAIF and Linux Foundation so that it can be neutrally governed and shared by all.
This is a pivotal moment for RL, agents, and the whole field, and making agents’ learning environments open source is only getting more We believe, now more than ever, that democratizing AI requires a fully open stack. Reinforcement Learning environments are a fundamental part of that stack because they teach agents the capabilities they will use in the real world.
The committee behind OpenEnv is proud to bring AAIF into Reinforcement Learning, and further down the stack, by sharing what we’ve built and used at; Hugging Face, Microsoft, Reflection, Meta-PyTorch, Fleet, Mercor, RadixArk, Nebius, Modal, Prime Intellect, Unsloth, and more.
This is a pivotal moment for RL, agents, and the whole field, and making agents’ learning environments open source is only getting more We believe, now more than ever, that democratizing AI requires a fully open stack. Reinforcement Learning environments are a fundamental part of that stack because they teach agents the capabilities they will use in the real world.
The committee behind OpenEnv is proud to bring AAIF into Reinforcement Learning, and further down the stack, by sharing what we’ve built and used at; Hugging Face, Microsoft, Reflection, Meta-PyTorch, Fleet, Mercor, RadixArk, Nebius, Modal, Prime Intellect, Unsloth, and more.
11
15
83
6,929
AK retweeted
We're announcing the public beta of datavendor.ai
We spent the last two years building the best solution for people buying and selling data, whether you're a researcher in a lab or a company trying to sell data.
Data is the fundamental bottleneck for AI progress, and infra that solves it.
We spent the last two years building the best solution for people buying and selling data, whether you're a researcher in a lab or a company trying to sell data.
Data is the fundamental bottleneck for AI progress, and infra that solves it.
216
156
2,005
631,809
AK retweeted
SpatialScout: Can VLM agents actively explore and map 3D scenes?
A new benchmark for evaluating vision-language models on active object-level 3D mapping across 80 scenes and 3,140 objects.
A new benchmark for evaluating vision-language models on active object-level 3D mapping across 80 scenes and 3,140 objects.
3
13
74
7,080
AK retweeted
Today distributed cognition exits out of stealth with backing by Verdict Capital.
Our mission is to replicate humanity's collective intelligence.
Read more: dcog.com/research/our-i…
Most of today's AI power is held in the hands of a few, who can choose what you're allowed to ermitted to do.
We stand on three substantially different foundational beliefs:
- You shouldn't ask permission for intelligence.
- Superintelligence will be a distributed collective.
- Humans are not meant to be relegated to a permanent underclass.
If we don't distribute intelligence, nobody will.
Our mission is to replicate humanity's collective intelligence.
Read more: dcog.com/research/our-i…
Most of today's AI power is held in the hands of a few, who can choose what you're allowed to ermitted to do.
We stand on three substantially different foundational beliefs:
- You shouldn't ask permission for intelligence.
- Superintelligence will be a distributed collective.
- Humans are not meant to be relegated to a permanent underclass.
If we don't distribute intelligence, nobody will.
188
109
568
1,052,377
AK retweeted
StreamEval: The first benchmark for interactive streaming video generation
It evaluates 37 models across 5 tasks with 300 scripted test cases, revealing that reliable request execution remains a central challenge.
It evaluates 37 models across 5 tasks with 300 scripted test cases, revealing that reliable request execution remains a central challenge.
5
10
41
5,785
AK retweeted
Predicting a stock's next move?
That's the warm-up act.
The real exam for LLM agents: building signals from raw data, learning from losses, and making money on data it has never seen before.
That's exactly what we built. Today we’re releasing Quant-Arena, a benchmark for research on real A-share market data.
9 frontier models. 3,042 stocks. One sealed year, 2024. Only 4 of the 9 models made money.Quant research is still wide open.
Full breakdown 👇
Project: lab.einsia.ai/quant-arena
Paper: arxiv.org/abs/2610.30113
Arena: quant.lab.einsia.ai/arena
That's the warm-up act.
The real exam for LLM agents: building signals from raw data, learning from losses, and making money on data it has never seen before.
That's exactly what we built. Today we’re releasing Quant-Arena, a benchmark for research on real A-share market data.
9 frontier models. 3,042 stocks. One sealed year, 2024. Only 4 of the 9 models made money.Quant research is still wide open.
Full breakdown 👇
Project: lab.einsia.ai/quant-arena
Paper: arxiv.org/abs/2610.30113
Arena: quant.lab.einsia.ai/arena
28
91
148
448,868
Looking Ahead
Modeling Future Evolution in Autoregressive Video Generation
paper: huggingface.co/papers/2610.29…
Modeling Future Evolution in Autoregressive Video Generation
paper: huggingface.co/papers/2610.29…
17
18
78
18,986
AK retweeted
Ropedia W0
A world model that learns loco-manipulation from 10k hours of human egocentric video, then transfers across embodiments, from robot arms to humanoids and from grippers to dexterous hands.
A world model that learns loco-manipulation from 10k hours of human egocentric video, then transfers across embodiments, from robot arms to humanoids and from grippers to dexterous hands.
7
12
37
5,543
AK retweeted
Introducing Flint, @Recursive_SI’s first model: a fast local coding model built with our automated research system, the Eureka Machine.
• Best-in-class fast local coding
• Up to 2x faster than comparable models: 125+ tokens/sec on a MacBook Pro, 150+ on RTX Spark
• Built from hs
Private, unmetered intelligence on your own machine. Open weights available now on @huggingface, run it locally on your favorite framework or try it for free on @opencode or @pidotdev.
• Best-in-class fast local coding
• Up to 2x faster than comparable models: 125+ tokens/sec on a MacBook Pro, 150+ on RTX Spark
• Built from hs
Private, unmetered intelligence on your own machine. Open weights available now on @huggingface, run it locally on your favorite framework or try it for free on @opencode or @pidotdev.
635
892
8,428
3,409,432
AK retweeted
👻 It's time to play with Ornith-1.5-122B on HF 🤗
Thanks @_akhaliq and @huggingface for the support!!!
🔗 huggingface.co/spaces/ornith-…
Thanks @_akhaliq and @huggingface for the support!!!
🔗 huggingface.co/spaces/ornith-…
Aloha!🌺 Meet Ornith-1.5-122B: No Tradeoff Between Strong and Fast
Stronger. Faster. Leaner. Ornith-1.5-122B delivers frontier-level capabilities with exceptional inference efficiency. With superior instruction following, stronger task decomposition, and deeper understanding , it achieves state-of-the-art performance among open-source models of comparable size, while running on a single NVIDIA DGX card!!
Open weights under the MIT license.
Coding
✅SWE-bench Verified: 0.845
✅TB2.1: 0.813
✅CyberGym: 0.800
✅NL2Repo: 0.512
Agentic
✅HLE: 0.446
✅BrowseComp: 0.785
✅MCP-Atlas: 0.746
✅Toolathlon: 0.708
🤗Hugging Face: huggingface.co/ornith-ai/Orni…
We can’t wait to see what you build with Ornith-1.5-122B.
6
10
66
7,680