#liquidai — Public Fediverse posts

🚀 Behold, the #LiquidAI #LFM2.5 model—is it half the size or half the IQ? 🤔 A mere 2.6B model that claims to punch in the 10B weight class, all while swimming in a sea of acronyms, emojis, and buzzwords. 🤯 Hugging Face seems more like a hugging maze of #jargon and enterprise pitches. 🎭
https://huggingface.co/LiquidAI/LFM2.5-2.6B #HuggingFace #AIbuzz #technology #HackerNews #ngated

🤖 #LiquidAI released #LFM2_5 2.6B, an #agentic model that runs entirely on-device: planning, tool calling & multi-step tasks without any cloud API #AI #LLM #EdgeAI #opensource
🧵👇

⚡ Decodes 220 tokens/s on an M5 Max CPU, 113 tokens/s on a Ryzen AI Max+ 395 and 30 tokens/s on a phone, staying under 2.5 GB memory

Reducing Doom Loops with Final Token Preference Optimization

https://www.liquid.ai/blog/antidoom

Comments: https://news.ycombinator.com/item?id=48820127

#HackerNews #DoomLoops #FinalTokenPreference #Optimization #AIResearch #LiquidAI

🍹 Introducing Liquid AI's latest cocktail: an "Even Better onDevice MixtureofExperts" with a splash of #buzzwords and a twist of jargon 🍹 Just what you needed! Now you can "unlock" your business potential by drowning it in a sea of acronyms and overpriced "solutions" that claim to be the world's "most efficient" 🙄 Cheers to that! 🎉
https://www.liquid.ai/blog/lfm2-5-8b-a1b #LiquidAI #Cocktail #BusinessSolutions #MixtureofExperts #OverpricedSolutions #HackerNews #ngated

Liquid AI reveals 8B-A1B MoE trained on 38T

https://www.liquid.ai/blog/lfm2-5-8b-a1b

#HackerNews #LiquidAI #MoE #AItechnology #8BA1B #machinelearning #38Tdata

Liquid AI has released LFM2.5-350M, a compact 350M parameter model trained on 28 trillion tokens that outperforms models more than twice its size. The model uses a hybrid LIV architecture supporting a 32k context window while maintaining a lean memory footprint. https://www.marktechpost.com/2026/03/31/liquid-ai-released-lfm2-5-350m-a-compact-350m-parameter-model-trained-on-28t-tokens-with-scaled-reinforcement-learning/ #AIagent #AI #GenAI #AIResearch #LiquidAI

Liquid AI releases LFM2-24B-A2B, a hybrid architecture that blends attention with convolutions to solve modern LLM scaling bottlenecks: Using a 1:3 ratio of attention to gated convolutions with sparse MoE, the model achieves 24B parameters while only activating 2.3B, fitting in 32GB RAM for local deployment. https://www.marktechpost.com/2026/02/25/liquid-ais-new-lfm2-24b-a2b-hybrid-architecture-blends-attention-with-convolutions-to-solve-the-scaling-bottlenecks-of-modern-llms/ #AIagent #AI #LLM #GenAI #LiquidAI

Sehr guter Text aus Ägypten zum Thema #Whistleblowing in den #USA speziell in der #AI Branche.

🌀 https://untoldmag.org/whistleblowers-ai-protection/

#openai #activism #surveillance #technology #law #internationallaw #google #liquidai

After receiving so many questions about my #zLNN and how #LiquidNeuralNetworks work, I’ve decided to start publishing articles to help people better understand LNNs. Check the first one out here:

https://tinyurl.com/mt242844

#AI #ArtificialIntelligence #MachineLearning #DeepLearning #LNN #LiquidAI #LiquidNeuralNetwork #Rust

MIT spinoff Liquid debuts small, efficient non-transformer AI models

Liquid AI, a startup by ex-MIT researchers, has launched its Liquid Foundation Models (LFMs), which outperform transformer-based models like Meta's Llama and Microsoft's Phi in speed and memory efficiency. A full launch event is set for October 23, 2024.

#LiquidAI #LFM #GPT #LLAMA #MicrosoftPhi #LLM #ArtificialIntelligence

https://venturebeat.com/ai/mit-spinoff-liquid-debuts-non-transformer-ai-models-and-theyre-already-state-of-the-art/

#LiquidAI, a new #MIT spinoff, wants to build an entirely new type of #AI https://techcrunch.com/2023/12/06/liquid-ai-a-new-mit-spinoff-wants-to-build-an-entirely-new-type-of-ai/