AtomicChat Strips LFM2.5 Refusals From 96% to 1% by Editing Later Blocks
AtomicChat shipped refusal-ablated GGUF builds of Liquid AI's LFM2.5-24B-A2B MoE, cutting English refusals from 96.3% to 1.2% while gaining MMLU points.
20 stories
AtomicChat shipped refusal-ablated GGUF builds of Liquid AI's LFM2.5-24B-A2B MoE, cutting English refusals from 96.3% to 1.2% while gaining MMLU points.
Liquid AI's d1.2 and d1.2-ultra top the Artificial Analysis Decision Index with audio input, five new structured output tasks, and near-perfect calibration.
Sarvam AI opens its India-hosted inference stack to everyone, undercutting global providers on price while matching them on speed across frontier open-weight models.
Sakana AI releases JFinBench, a 321-question Japanese finance benchmark built from 2026 certification exams that exposes a persistent gap between open-weight and proprietary models.
Tencent's Hy4 flash packs 224B total parameters with 12B active, matches Claude Haiku 5.5 on daily tasks at 1/18th the cost
A new benchmark tests whether vision-language models can infer culture, period, origin, and creator from museum images, and they mostly cannot.
Alibaba's new flagship uses a fresh MoE design to cut token usage nearly in half while beating Qwen3.8-Max on productivity benchmarks.
A new open-source orchestration layer from HKU lets multiple coding agents like Codex, Claude Code, and Cursor coordinate on one task.
ENIGMA-360 pairs 180 headset videos with 180 synced fixed-camera videos of real repair work. Top video models break down when they have to switch between the two views.
Veyra is an open source Windows tool that pipes PS5, Xbox, capture cards and local video through DLSS, FSR and AI denoisers in real time.
Databricks now lets teams define business metrics once in Unity Catalog and publish them directly to Power BI semantic models, keeping definitions in sync.
A local-first Electron app translates English PDFs into side-by-side Chinese, pairs every paragraph with an AI chat, and plugs into Claude Code or Codex for annotations.
A GGUF-packaged Whisper large-v3 built for transcribe.cpp hits 1.81% WER on LibriSpeech and runs on a laptop in under 1 GB.
A community-trained LoRA for Krea 2 focused on Asian character generation has crossed one million downloads, pulling the open-weights base model into the spotlight.
Osaurus released Diplodocus preview 4 as llama.cpp GGUF files, a 9B finance and spreadsheet assistant built on K2-Horizon-7B with tool use.
A self-hosted Slack clone where Claude, Codex, Gemini and Grok agents join channels as teammates, with approvals, budgets and MCP tools
A from-scratch Vietnamese TTS hits 48 kHz audio, instant voice cloning, and real-time CPU inference with an OpenAI-compatible streaming API.
A new Transformer layout runs full blocks side by side. In small MoE tests it beats standard stacks on quality at matched compute and decodes up to 51% faster.
ChatR Labs ships a rank-128 LoRA merged into Qwen-Image-2.1 across GGUF, FP8, INT8, NVFP4 and NF4 builds, with an uncensored tune that improves adult-prompt adherence while running on 4 to 14 GB of VRAM.
FedBNN trains one-bit neural networks across federated devices using rotation-aware binarization, cutting FLOPs up to 58x and memory 32x with competitive accuracy.