Records found: 5
'Liquid AI launched LFM2-VL-3B, a 3B parameter vision language model built for edge deployments with efficient token compression, native aspect ratio support, and documented inference controls for predictable latency'
'Liquid AI released LFM2-8B-A1B, an 8.3B-parameter sparse Mixture-of-Experts designed for phones and edge devices that activates about 1.5B parameters per token to balance latency and capacity.'
'Neuphonic released NeuTTS Air, a 748M-parameter on-device TTS model that clones voices from ~3 seconds of reference audio and runs locally via GGUF quantizations for CPU-first real-time synthesis.'
IBM released Granite 4.0, a hybrid Mamba-2/Transformer LLM family that cuts serving memory by over 70% for long-context inference while keeping strong instruction-following and tool-use performance.
'A concise 2025 guide to the top local LLMs, comparing context lengths, VRAM needs, quant presets, and license implications for practical on-prem deployments.'