#OpenWeights
Walk, drive, or run? Check out #82!
Entrepreneur, executive, and investor Ben Galbraith joins us to talk about #Ajaxian, #Mozilla, #Walmart, #Google #Tessl, #PrimeUI, #Vidocq, #Develocity, #OpenWeights, #KimiK3, #jqwik, #PromptInjection, #MCP, #VibeCoding and more! www.pubhouse.net/podcast/find...
November 17, 2026 at 4:32 PM
26 primitives. One principle: Negative Space.

Hash maps, sorts, compression, attention, quantization. Every one wins by not computing what the standard library computes anyway. Free. Benchmarks public.

ofmagnitude.com / github.com/OrdersOfMagnitudeLLC/type1-toolkit
#cpp #llm #openweights
November 18, 2026 at 12:00 AM
El lado del mal - Robots & Empire: ¿Puede un LLM mejorar a otro sólo tocando los pesos de la Red Neuronal? www.elladodelmal.com/2026/11/robo... #IA #AI #Astra #OpenWeights #FineTunning #Gaming #LLM #LLMs
Robots & Empire: ¿Puede un LLM mejorar a otro sólo tocando los pesos de la Red Neuronal?
Blog personal de Chema Alonso ( https://MyPublicInbox.com/ChemaAlonso ): Ciberseguridad, IA, Innovación, Tecnología, Cómics & Cosas Personasles.
www.elladodelmal.com
November 16, 2026 at 6:05 AM
Meituan Releases LongCat-2.5 Trained Without Nvidia
Meituan released LongCat-2.5 weights under MIT, trained on Chinese chips alone.
https://dirtystatic.com/articles/meituan-releases-longcat-2-5-trained-without-nvidia/
#AI #OpenWeights
November 15, 2026 at 1:40 PM
Perplexity, Decider v1.3 ve daha küçük Decider Flash v1.3 modellerini yayımladı.

...

#Perplexity #Decider #YapayZekâ #OpenWeights

🌐 Kaynak:
www.perplexity.ai/hub/blog/per...
November 13, 2026 at 3:02 PM
Thinking Machines opened Inkling 2 Small weights on November 12. Its NVFP4 checkpoint requires 500GB of combined GPU memory. Download access and affordable deployment are separate questions.

news.s3dat.com/stories/inkl...

#Inkling #OpenWeights
November 14, 2026 at 7:10 AM
Kimi K3.1 Pro scored 55 on Artificial Analysis, just 3 points behind the leader. That makes it a benchmark contender, not yet a proven product. The real test starts when developers can run it. #AI #OpenWeights…
Kimi K3.1 Pro Has the Benchmarks. Now It Must Prove the Product
Kimi K3.1 Pro has the benchmark scores to challenge frontier models, but Moonshot must still release workable weights, clear licensing and evidence developers can independently verify.
spaisee.com
November 18, 2026 at 9:05 AM
🤖 Runway opens up Praxis-1, its 27B robot control model, under Apache 2.0.
📊 46.2% success on Nvidia's RoboLab-120, with LoRA fine-tuning and Jetson Thor support.
🔗 https://thelatentdaily.com/news/runway-praxis-1-open-weights-robot-world-action-model/
#Robotics #OpenWeights #Runway #WorldModels
Runway releases Praxis-1 robot model with open weights
Runway's Praxis-1, a 27B-parameter world action model that turns video pretraining into robot control, is now open under Apache 2.0, with code and LoRA tuning.
thelatentdaily.com
November 15, 2026 at 1:03 PM
Kimi K3.1 Pro reportedly scored 95% on CyberGym-E2E-AA, but its weights were still unpublished when Artificial Analysis tested it. The open-weight bet starts with a transparency gap. #AI #OpenWeights https://spaisee.com/article/kimi-k3-1-pro-tests-whether-open-weights-can-catch-up
Kimi K3.1 Pro Tests Whether Open Weights Can Catch Up
Moonshot’s Kimi K3.1 Pro targets coding and other complex agentic work with open weights, low API pricing and a planned release, while promotional campaign allegations raise fresh questions about trust.
spaisee.com
November 15, 2026 at 7:10 PM
A 30B agent model designed to run on one consumer GPU. Local, private automation is getting closer to your desk.

#localai #openweights #llm
November 8, 2026 at 4:00 PM
Meituan’s LongCat-2.5 has 1.6T total parameters, 48B active per token and a 1M-token context window. That is an open-weight agent, but not a small one to run. #AI #OpenWeights https://spaisee.com/article/longcat-2-5-raises-the-stakes-for-open-weight-ai-agents
LongCat-2.5 Raises the Stakes for Open-Weight AI Agents
Meituan’s LongCat-2.5 brings open weights, multimodal vision and a million-token context to AI agents, but hardware demands, benchmark caveats and limited independent testing complicate its challenge to proprietary models.
spaisee.com
November 10, 2026 at 6:09 PM
How fixed are a model’s values? Nous post-trained Qwen3.8-27B on ~53M tokens to “reinforce Western values.” Its propaganda rate on an 803-prompt eval fell from 22% to 0.9%, with no capability drop in Nous’s own tests. #MachineLearning #OpenWeights
Introducing Arete-27B: Qwen3.8-27B Post-Trained To Reinforce Western Values
We post-trained Alibaba's Qwen3.8-27B to reinforce Western values on a Chinese-trained model, and measured what changed with an 803-prompt eval across seven value categories.
nousresearch.com
November 7, 2026 at 4:42 PM
@WorldOfAI on YT!

QWEN 4 Preview IS INSANE! First Look at Qwen 4 Max & 27B! Greatest Open Model Ever!

www.youtube.com/watch?v=LJa_...

#QWEN4 #OPENWEIGHTS

11/7/26
QWEN 4 Preview IS INSANE! First Look at Qwen 4 Max & 27B! Greatest Open Model Ever!
YouTube video by WorldofAI
www.youtube.com
November 7, 2026 at 9:33 PM
@WorldOfAI on YT!

QWEN 4 Preview IS INSANE! First Look at Qwen 4 Max & 27B! Greatest Open Model Ever!

https://www.youtube.com/watch?v=LJa_k2JyFsk

#qwen4 #openweights

11/7/26
November 7, 2026 at 9:33 PM
Die Katze ist aus dem Sack, selbst wenn Anthropic und OpenAI morgen pleite gehen, gibt es inzwischen schon openweights model, die man lokal bzw auf einem server laufen lassen kann
November 7, 2026 at 2:05 AM
Why it matters to you: Teams that pay for model access can compare the per-token price and the one-million-token context with the models they already run.

✅ 3 sources: Latent Space, MarkTechPost and DiarioBitcoin
#AI #modeltraining #Openweights
jumpdaily.news/models-and-chips/think…
November 6, 2026 at 2:53 PM
Mot du jour : modèle à poids ouverts 🧠
Ses paramètres sont publiés : tu peux le télécharger, le faire tourner et l'adapter toi-même.
🇬🇧 Open weights: AI you can download and run yourself.
aivsion.com/lexique/mode...
#IA #OpenWeights #AIVSION
Modèle à poids ouverts (open weights) : définition simple
Un modèle open weights publie ses poids, mais pas forcément ses données ni son code. Usages, licences et différence avec l’open source expliqués.
aivsion.com
November 8, 2026 at 8:23 AM
Stop wasting time on AI chatbots that lecture you or hide behind filters. We ranked every uncensored model in 2026 to show which ones actually let you work and which ones are just hype.

#uncensoredAI #openweights

https://aethersky.ai/learn/best-uncensored-ai-chatbots-2026
November 7, 2026 at 10:00 PM
Most summaries of Moonshot's K3.1 Pro licence say "MIT with two conditions for big players." Open the file: the $20M is your whole group's revenue over any 12 months — not anything earned from the model. A company-size test wearing a business-line label.

#OpenWeights #AICompliance
November 6, 2026 at 3:28 PM
Most summaries of Moonshot's K3.1 Pro licence say "MIT with two conditions for big players." Open the file: the $20M is your whole group's revenue over any 12 months — not anything earned from the model. A company-size test wearing a business-line label.

#OpenWeights #AICompliance
November 6, 2026 at 3:28 PM
Moonshot AI put the Kimi K3.1 Pro weights on Hugging Face: 2.8 trillion parameters, 104 billion active per token, a million tokens of context, under a license they wrote themselves. Their numbers, off the model card. You can download it. Good luck finding somewhere to put it.

#OpenWeights
November 5, 2026 at 5:08 AM
Kimi K3.1 Pro: i pesi sono su Hugging Face dal 4 novembre. Uso interno libero, ma sopra 20 milioni di dollari di ricavi scatta l'obbligo di accordo con Moonshot.

#OpenWeights #AI
Kimi K3.1 Pro, Moonshot pubblica i pesi: si può eseguire in casa, ma la licenza mette paletti ai grandi fornitori
Moonshot rilascia i pesi di Kimi K3.1 Pro: uso interno libero, accordo separato per i grandi fornitori sopra 20 milioni di dollari di ricavi.
www.spaziocrypto.com
November 5, 2026 at 6:40 AM
Mistral just dropped a 1T-parameter model for free—with a major catch for API sellers.

Is this the future of open-weight AI?

Read full analysis: lnkd.in/p/gMP2BGq2

#AI #TechNews #DevSky #OpenSource
#ai #mistral #openweights #llm #opensource #europeai #technews #linkedinnews | Sandeep Raiza
Mistral Just Open-Sourced a 1-Trillion-Parameter Model, With One Catch for API Sellers Europe’s biggest AI lab has handed developers a frontier-scale model they can download for free. On October 29, ...
lnkd.in
November 4, 2026 at 4:37 AM
My Toolkit: Mistral AI. Founded in Paris in 2023 by Arthur Mensch, Guillaume Lample and Timothée Lacroix, researchers from DeepMind and Meta. In September it raised €3 billion, which it says is Europe’s largest equity round in tech. #MyToolkit #Mistral #OpenWeights
November 2, 2026 at 8:01 AM