Engines and Tools

ComfyUI v0.40.0 Released: Native LoRA Stack and More

ComfyUI v0.40.0 is released with a native LoRA stack loader, Qwen 3.5 LLM low-VRAM optimizations, and expanded partne ...

New Models

LFM2.5-24B-A2B-MLX-6bit Text Generation Model: 24GB+ VRAM

Liquid AI has released a 6-bit MLX quantized version of LFM2.5-24B-A2B, optimized for fast and private on-device agen ...

New Models

LFM2.5-24B-A2B Agent-Optimized MoE Reasoning Model: 12GB+ VRAM

LiquidAI has released LFM2.5-24B-A2B, a 24B reasoning model with 2.45B active parameters optimized for fast, private ...

New Models

LFM2.5-24B-A2B-MLX-8bit Text Generation Model: 32GB+ VRAM

LiquidAI’s LFM2.5-24B-A2B reasoning MoE model is now available in MLX 8bit format for Apple Silicon, optimized ...

New Models

LFM2.5-24B-A2B-MLX-bf16 Reasoning Model for Apple Silicon: 80GB+ VRAM

LiquidAI has released a bfloat16 MLX export of the LFM2.5-24B-A2B reasoning model, optimized for Apple Silicon device ...

New Models

LiquidAI Releases LFM2.5-24B-A2B for On-Device Agents

LiquidAI has released LFM2.5-24B-A2B, a 24B MoE reasoning model with 2.45B active parameters optimized for fast, priv ...

New Models

LFM2.5-24B-A2B-FP8 Text Generation Model: 32GB+ VRAM

Liquid AI has released LiquidAI/LFM2.5-24B-A2B-FP8, an FP8 quantized version of their inference model designed for ef ...

Engines and Tools

Unsloth v0.1.907-beta Released with Major Speed and UI Upgrades

Unsloth v0.1.907-beta is out, bringing up to 10x faster UI responsiveness, 5x faster inference, new video models, and ...

Engines and Tools

Magnitude CLI 0.2.8 Released with VRAM Fixes and Gemma 4 Support

Magnitude v0.2.8 is out, featuring 8 GB GPU VRAM management improvements, Gemma 4 draft model support, embedding stre ...

Engines and Tools

KoboldCpp v1.123 Released: Major Agent Feature Upgrades

KoboldCpp v1.123 is now available, featuring major agent enhancements, parallel request mode overhaul, Qwen Image 2.1 ...