Thom_Wolf Profile Banner
Thomas Wolf Profile
Thomas Wolf

@Thom_Wolf

Followers
91K
Following
27K
Media
465
Statuses
5K

Co-founder at @HuggingFace - open-source and open-science

Joined February 2011
Don't wanna be here? Send us removal request.
@Thom_Wolf
Thomas Wolf
8 days
We've had so much fun organizing a vibe coding event for kids aged 9 to 13. that we decided to show our friends how to host one too. and turn it into a global event. 📆 It’s happening on October 10-11, 2025.🧑‍🏫 with in-person sessions hosted locally,.🌐 all connected in one giant
20
41
188
@Thom_Wolf
Thomas Wolf
15 hours
Tweet media one
@Thom_Wolf
Thomas Wolf
22 hours
Thrilled to finally share what we've been working on for months at @huggingface 🤝@pollenrobotics. Our first robot: Reachy Mini. A dream come true: cute and low priced, hackable yet easy to use, powered by open-source and the infinite community. Tiny price, small size, huge
3
8
40
@Thom_Wolf
Thomas Wolf
16 hours
RT @betterhn50: Hugging Face just launched a $299 robot that could disrupt the robotics industry (.
0
2
0
@Thom_Wolf
Thomas Wolf
17 hours
released 5h ago – just passed €130 000 in pre-order
Tweet media one
@Thom_Wolf
Thomas Wolf
22 hours
Thrilled to finally share what we've been working on for months at @huggingface 🤝@pollenrobotics. Our first robot: Reachy Mini. A dream come true: cute and low priced, hackable yet easy to use, powered by open-source and the infinite community. Tiny price, small size, huge
25
13
386
@Thom_Wolf
Thomas Wolf
17 hours
when your multiple recent releases cross each others on the front page of reddit
Tweet media one
5
3
39
@Thom_Wolf
Thomas Wolf
17 hours
RT @iamwill: cute compute…❤️❤️❤️.
0
37
0
@Thom_Wolf
Thomas Wolf
19 hours
sorry Jony
Tweet media one
@Thom_Wolf
Thomas Wolf
22 hours
Thrilled to finally share what we've been working on for months at @huggingface 🤝@pollenrobotics. Our first robot: Reachy Mini. A dream come true: cute and low priced, hackable yet easy to use, powered by open-source and the infinite community. Tiny price, small size, huge
1
2
35
@Thom_Wolf
Thomas Wolf
20 hours
your turn @Apple.
@Rufus87078959
Rufus
20 hours
If Apple had launched this, they would have increased their market cap by $1T, but that was the old Apple, not the current one. A huge release from Hugging Face; I hope they can build on this launch. They shouldn't lose this momentum. @ClementDelangue.
3
3
32
@Thom_Wolf
Thomas Wolf
20 hours
RT @hypermaddd: Prepare new world. New meta.
0
7
0
@Thom_Wolf
Thomas Wolf
20 hours
RT @JeremyNguyenPhD: Cute, open-source robot you can program in python!.
0
4
0
@Thom_Wolf
Thomas Wolf
21 hours
RT @GoingBallistic5: The current generation of roboticists grew-up on Lego Mindstorm. The Reachy Mini is for the next generation. Bravo!. B….
0
7
0
@Thom_Wolf
Thomas Wolf
21 hours
RT @MartinSignoux: 👏this👏is👏so👏cool👏.
0
6
0
@Thom_Wolf
Thomas Wolf
21 hours
RT @csgbwk: I really love the design, exactly whatI'm expecting from a desk assistant, really reminds me of ELEGNT by Apple and i love that….
0
2
0
@Thom_Wolf
Thomas Wolf
22 hours
Thrilled to finally share what we've been working on for months at @huggingface 🤝@pollenrobotics. Our first robot: Reachy Mini. A dream come true: cute and low priced, hackable yet easy to use, powered by open-source and the infinite community. Tiny price, small size, huge
209
435
3K
@Thom_Wolf
Thomas Wolf
24 hours
And all the details are in the tech report blog:
@Thom_Wolf
Thomas Wolf
24 hours
A TLDR thread 🧵.
0
0
4
@Thom_Wolf
Thomas Wolf
24 hours
the result is a solid model which aim at checking all the boxes:.- small – /to run on the edge/.- fast inference – /with GQA/.- multilingual – /with coverage for 6 languages/.- long-context – /for document editing and tool use/.- reasoning – /with optional.
4
0
9
@Thom_Wolf
Thomas Wolf
24 hours
and a final extensive step of model-merging to recover and balance all behaviors
Tweet media one
1
0
5
@Thom_Wolf
Thomas Wolf
24 hours
before post-training for alignement/instruction-following using Anchored Preference Optimization (APO) – the latest variant of DPO
Tweet media one
1
0
3
@Thom_Wolf
Thomas Wolf
24 hours
with a new mid-training step for long-context extension and reasoning
Tweet media one
1
0
3
@Thom_Wolf
Thomas Wolf
24 hours
we even created new datasets for pre-training and fine-tuning - extensive ablations for the best data mixture
Tweet media one
2
0
6
@Thom_Wolf
Thomas Wolf
24 hours
we scaled it to train for 24 day on 384 H100 GPU
Tweet media one
1
0
4