lfm25

Local LLM, image&video&music generator, run codex with local models on your phone

LFM2-powered research agent | Automated Related Work & citations | Semantic Scholar + arXiv

LFM2.5 1.2 billion reasoning language model runs on web browsers using Transformers.js and ONNX Runtime Web.

LFM2.5 1.2 billion Japanese language model runs on web browsers using Transformers.js and ONNX Runtime Web.

A minimal implementation of Liquid Foundation Model LFM 2.5 2.6B model

A dependency-free C inference engine for Liquid AI's LFM2 / LFM2.5 models. Terminal chat, an OpenAI-compatible server, a web agent and a terminal agent, all in one executable.

Building a Multimodal Model with LFM2.5 and Kimi-K2.6

A self-contained CUDA inference engine for LiquidAI/LFM2.5-8B-A1B (hybrid conv + GQA-attention MoE, 8.5B params, 1B active) targeting a single RTX 3060 (12 GB). No Python, no frameworks at runtime: a single .cu engine + a header-only byte-level BPE tokenizer.

On-device iOS assistant: a QLoRA-tuned LFM2.5-2.6B tool-calling LLM runs on the iPhone via Apple MLX and drives real system actions.

Python script to convert a document (PDF, EPUB, XPS, MOBI, FB2, CBZ, images, ... - anything PyMuPDF can open) into readable UTF-8 HTML, Markdown, or plain text, extracting embedded images along the way.

Same prompt, all 22 Turbo Brilliance reasoning modes, one small local model on one GPU: pick your reasoning effort with measured numbers, not vibes

Looped hybrid RAG lab about the cost of a second look: local models, visible traces, published metrics.