Liquid AI released LFM2.5-8B-A1B with a tokenizer expanded from 65K to 128K vocabulary, doubling the size to improve tokenization for Hindi, Vietnamese, and Thai without retraining from scratch. The n…
Nativ has released an open-source desktop app that lets users run frontier open models from Google, Cohere, and Liquid AI locally on Apple Silicon Macs with no accounts, subscriptions, or cloud depend…
Liquid AI open-sourced Antidoom, a Final Token Preference Optimization method that reduces doom loops in reasoning models by retraining only the first loop-start token. On LFM2.5-2.6B, looping fell fr…
Researchers at Liquid AI developed Antidoom, a method using Final Token Preference Optimization (FTPO) to reduce 'doom loops'—repetitive inference failures in small reasoning models—from 10.2% to 1.4%…
NobodyWho released NobodyWho Chat, a fully offline AI assistant for iOS and Android that runs open-weight models locally on the device, ensuring no data leaves the phone and no internet connection is …
Liquid AI released LFM2.5-230M, a 230-million-parameter model optimized for edge devices including phones, Raspberry Pi, and robots. The model achieves 213 tokens per second on a Galaxy S25 Ultra and …
Liquid AI released LFM2.5-230M, its smallest open-weight model optimized for on-device agentic tasks like data extraction and tool use, achieving 213 tok/s on a Galaxy S25 Ultra and 42 tok/s on a Rasp…
Liquid AI released LFM 2.5-230M, a 230-million-parameter model that outperforms larger models on data extraction benchmarks, achieving 22.51 on CaseReportBench versus 13.83 for Qwen3.5-0.8B and 2.28 f…
Liquid AI released its LFM2.5-230M foundation model on June 25, outperforming larger competitors like Alibaba's Qwen3.5-0.8B and Google's Gemma 3 1B on data extraction benchmarks. The 230-million-para…
Liquid AI released two new retrieval models, LFM2.5-Embedding-350M and LFM2.5-ColBERT-350M, for fast multilingual search across 11 languages. The models, available on Hugging Face, are designed as dro…
Squeez Labs demonstrated CrankGPT, a fully offline, hand-cranked AI voice assistant built on a Raspberry Pi 5 with 8GB RAM and a 20W generator, running local LLMs via llama.cpp. The device boots in 30…
NobodyWho has released the world's first standalone Apple Watch app that runs AI models locally and offline without requiring an iOS companion app. The app supports small language models like Prism ML…
Liquid AI's LFM2 model can be fine-tuned using QLoRA and DPO in a complete open-source workflow on Google Colab. The process loads the base LFM2 checkpoint with 4-bit quantization, trains a lightweigh…
Liquid AI released LFM2.5-8B-A1B, an edge model designed for fast tool calling on consumer hardware, with a 128K context window and pretraining scaled to 38 trillion tokens. The model, available on Hu…
Liquid AI released LFM2.5-8B-A1B, an on-device Mixture-of-Experts model with 8.3 billion total parameters and 1.5 billion active parameters per token, designed for tool calling and reasoning on consum…
Apple plans to make on-device artificial intelligence the centerpiece of its Worldwide Developers Conference next month, leveraging custom silicon to process AI queries locally as a privacy-focused al…