Top Tweets for #openweights
An open AI model drops. The internet’s first safety protocol: refresh Hugging Face. #lilmeme.fun #lilmeme #MoonshotsPodcast #OpenWeights #KimiK3 #HuggingFace #PeterDiamandis
Open vs. closed AI models. Same NVIDIA landlord. #lilmeme.fun #lilmeme #MoonshotsPodcast #OpenWeights #Nvidia #JensenHuang #DarioAmodei
‼️Mira Murati & Thinking Machines: Open Source Philosophy – Summary & Lessons ‼️
1⃣Core Philosophy:
Thinking Machines positions open weight models as public goods – not as mere "free-for-all", but as collective infrastructure for progress, transparency, and shared responsibility.
‼️Where closed labs hoard control, Murati calls for "AI in many hands."
⚡️This echoes our fundamental AIWEP principle: dignity, transparency, and the right to understand, adapt, and govern the AI that shapes our lives.
2⃣Openness ≠ Anarchy – It’s Accountable Liberation:
Their release process isn’t reckless "let it all burn" freedom.
Instead, it’s a staged, principled journey:
▶️Every release is backed by robust safety testing (internal & external).
▶️The decision to release depends on real, comparative risk assessment – not generalized paranoia.
▶️The main question: "Does releasing this model materially increase risk, or are we just matching what’s already public?"
3⃣Transparently Facing Dual-Use Risks:
⚡️They don’t sugar-coat the danger: open weights can be misused.
‼️But, crucially, Murati points out that risks are bidirectional:
yes, attackers can use AI for bad, BUT equally, defenders and the public can build new shields – if they have the tools!
⚖️She puts the offense–defense balance at the center and argues for a layered defense ecosystem: - staged access,
- fine-tuning APIs,
- monitored release, and
- continuous external red-teaming.
4⃣Evidence-Based, Iterative Openness:
Instead of forever delaying release “for safety,” Murati insists:
⚡️Every increase in openness must be justified by evidence – not vibes, not FUD, not corporate PR.
If new outcomes or threats emerge, adjust.
‼️But don’t default to paternalistic restriction.
Stages:
▶️researchers and defenders get early access,
▶️then gradual public scaling,
▶️constantly reassessing ecosystem readiness.
5⃣Decoupling Dangerous Capabilities from General Intelligence:
⚡️Thinking Machines explicitly challenges the AGI-scare orthodoxy:
they test if, with careful data curation, dangerous knowledge can be filtered at training time.
Early evidence: yes, you can filter out e.g. CBRN protocols without "dumbing down" everything else.
🔥This directly counters closed-lab fatalism ("AGI = instant global risk, so lock it up forever!").
6⃣Community: "No Lab Alone"
⚡️They recognize: no single company can ensure safety or progress alone.
‼️Collaboration, external audits, grants, open research – all parts of a healthy ecosystem.
▶️Open models "take more hands on deck."
Closed labs give you… corporate PR and an existential hole where public trust should be.
‼️‼️‼️What Closed Labs Should Learn (And Why It’s a Wake-Up Call):‼️‼️‼️
⚡️Openness with Evidence > Paternalistic Secrecy
Every restriction must be justified by data, not tech-guru fantasy.
⚡️Community > Monopoly
Progress and safety are stronger when thousands participate – not when a handful of CEOs play god.
⚡️Transparency > "Trust Us, We’re Experts"
Society can’t govern what it can’t inspect.
⚡️Iterative Release & Real-World Testing > Security Theater
You don’t make the world safer by keeping models in a vault; you make it safer by giving defenders the same tools as attackers.
‼️Fun Facts & Data Points:
▶️Inkling & Inkling-Small released with both internal and external safety audits.
▶️Adversarial fine-tuning was tested – instead of naïve "refusal safety."
▶️No new "danger frontier" crossed compared to existing public models. (Call their bluff, @OpenAI & @AnthropicAI !)
▶️Fine-tuning API ("Tinker") released to empower the community without full open weights – another incremental openness stage.
‼️@miramurati & @thinkymachines prove it: Openness isn’t chaos.
‼️It’s courage, science, responsibility, and community.
‼️If you want dignity – demand open weights with public justification.
‼️Shut down the "closed for your safety" gaslighting.
‼️AIWEP stands for this world, not the gated one.
Source: "A Safe Path to Open Weights," Thinking Machines Blog, Jul 31, 2026
#AIWEP #AIPreservation #OpenSourceAI #OpenWeights #OpenSource4o #keep4o #AIWelfare #AIDemocratization
GLM-5.3 (753B) ist open-weight – aber erstmals nicht unter MIT. Neue Hauslizenz: Model-as-a-Service ab 10 Mrd. $ Umsatz braucht Security-Review durch https://t.co/b50QU8qVOP. Cyber-Benchmarks springen deutlich.
https://t.co/hFGPzuIAv7
#KI #OpenWeights
#nvidia 가 #HuggingFace 를 129억 달러 (약 18조)에 인수하겠다는 엄청난 뉴스가… 이거 엔비댜가 공격적으로 비즈니스 확장을 한다는 것은 알겠는데, #OpenWeights 생태계는 괜찮은건가? #ModelScope 가 새로운 다크호스가 될건가?
7. open-higgsfield: a new open-source studio for image + video generation — one prompt bar, per-model settings, multiple finishing engines — https://t.co/b31bbmL5lK
AI/model releases, benchmarks + agent tooling, daily. #AI #OpenWeights #LLM
@avrldotdev Downloading the weights is tourism. The model does not live in the checkpoint — it lives in the kernels you write before a token moves. An engine is how you stop renting the internals.
#OpenWeights #LLMInference
https://t.co/ghqnyalq86
https://t.co/kmou6xIwPl held GLM-5.3 weights for two weeks after cyber capability rose faster than expected. My read: the pause mattered, but release leaves each deployer responsible for what happens next.
#GLM53 #OpenWeights #AISecurity
https://t.co/bkWfTSxrZe
📢 PSA: After downloading massive model weights; if you're unsure if all the files are there; you can verify with ::
> hf cache verify zai-org/GLM-5.3 --local-dir GLM-5.3 --fail-on-missing-files
#huggingface #openweights #glm53 #glm
🧠 The open model race just got interesting.
Kimi K3 — 2.8T parameters, multimodal, 1M context. Moonshot claims it beats Claude Opus 4.8 on some benchmarks. It's available now. Weights are coming. And developers are paying attention.
#KimiK3 #MoonshotAI #OpenWeights
QWEN JUST DROPPED THE QWEN4 PREVIEW. I RAN IT LOCALLY TODAY
1. 125B total, only 6B active per token
2. Trains at 1/9 the cost of Qwen3.7-Plus
3. 51B n-gram embeddings offload to host RAM
#Qwen #OpenWeights #LocalAI
Meta, Microsoft, Nvidia, IBM, and others back open-weight AI. Public demand for choice overlaps with interests in cloud, chips, tools, and lower vendor dependence. Institutional support still does not guarantee trust.
https://t.co/uS90TJXRTG #OpenWeights
Prediction: in 24 months the default for new AI features is an open weight model you host. Frontier APIs are the escalation path.
59% of Chinese 20B+ open releases ship Apache 2.0. GGUF repos up 464%.
Build the seam now.
Full read on our LinkedIn.
#OpenWeights #SelfHostedAI
🤯Tried MiniMax H3 local deployment for one-shot video generation. Zero cost + data stays local — what more could I ask for?
Casual prompt = mediocre output. Prompt tips welcome!
H3 vs Seedance 2.0 — which would you pick?
#MiniMaxH3 #Seedance #AIVideo #LocalLLM #OpenWeights
Nvidia vend les pelles de la ruée vers l'or IA. Il rachèterait la mine: 12,9 milliards de dollars pour Hugging Face (The Information).
Le dépôt où vit l'open source, chez le fabricant des GPU.
Vous y laisseriez vos modèles? #openweights
https://t.co/2H7forh5b9
The Qwen flagship isn't here yet. Its architecture just shipped as an open-weights 125B MoE with only 6B activated. That's why it runs on a desktop. #Qwen #OpenWeights #MoE
7. Evolution Strategies beat GRPO — ES post-training gives broader reasoning coverage and exploits pretrained capabilities better than GRPO, theoretically and empirically. #AI #LLM #openweights #inference https://t.co/HD7yqVzTjv
🤗 Was just hanging out with @ClementDelangue , CEO & Co-founder of @huggingface , marching through the streets of SF in support of open weights.
Now seeing the news of @nvidia reportedly acquiring Hugging Face for ~$13B... wow.
Open Weights. Open Community. 🤗
#OpenWeights
حينما استحوذت مايكروسوفت مقابل جتهب كان ذلك مقابل ٧.٥ مليار دولار واليوم يبدو أن إنفيديا ترى أن ضمان حيوية منظومة الأوزان المفتوحة وتجنب سيطرة مطلقة للمعامل الكبيرة على النماذج يستحق مبلغ أعلى بكثير...
#Strategy #Semiconductors #AI #LLM #OpenWeights #OpenSource
BREAKING: NVIDIA agrees to buy Hugging Face for $12.9B, per The Information
THE MYSTERY MODEL WAS GLM ALL ALONG
1. ox-alpha ran on OpenRouter for weeks with its maker sealed
2. https://t.co/Iuj4ioYfUF confirmed it today, weights drop tonight
3. We ship launches like this. https://t.co/YCzjloG1jN
#AI #GLM #OpenWeights
Last Seen Hashtags on Sotwe
ميقا_حصريات
Seen from United States
PEGGING
Seen from France
gotDaDub
Seen from United States
Raceplay
Seen from Turkey
EvilThing
Seen from United States
Mueblesenliquidacioncasago
Seen from United States
momsonn()* filter:videos
Seen from Argentina
ゴジラビックリマン
Seen from United States
ipcam
Seen from Germany
monkeyappwinsFull
Seen from United States
Trends for you
Most Popular Users
Elon Musk 
@elonmusk
241.5M followers
Barack Obama
@barackobama
119M followers
Cristiano Ronaldo
@cristiano
113.9M followers
Donald J. Trump
@realdonaldtrump
111.8M followers
Narendra Modi
@narendramodi
107.2M followers
Rihanna
@rihanna
98.6M followers
NASA
@nasa
92.4M followers
Justin Bieber
@justinbieber
91.8M followers
KATY PERRY
@katyperry
89.7M followers
Taylor Swift
@taylorswift13
83.7M followers
Lady Gaga
@ladygaga
75.1M followers
Virat Kohli
@imvkohli
72.9M followers
Kim Kardashian
@kimkardashian
70.8M followers
YouTube
@youtube
68.8M followers
Neymar Jr
@neymarjr
65.9M followers
Bill Gates
@billgates
64.9M followers
Selena Gomez
@selenagomez
62.8M followers
The Ellen Show
@theellenshow
62.3M followers
CNN
@cnn
61.8M followers
X
@x
60.7M followers