SoMark's OXR-1.0 Tops 3 Benchmarks Parsing PDFs 15x Faster
SoMark's 1.3B open-source parser tops three benchmarks, runs 15x faster than rivals, and uses a self-improvement loop to patch its own blind spots.
20 stories
SoMark's 1.3B open-source parser tops three benchmarks, runs 15x faster than rivals, and uses a self-improvement loop to patch its own blind spots.
Cloudflare released Clef and Clef-flash, open-source multimodal decision models that return calibrated probabilities over typed schemas instead of free-form text.
A new 8-step DMAD LoRA raises the quality ceiling on MiniMax-H3 text-to-audio-video, keeping 8 GB GPU support and ComfyUI drop-in workflows.
Perplexity shipped an open-weight 27B decision model and a hosted API that returns calibrated probabilities over your options instead of free text.
Fermion Research shipped a 164 MB English speech recognition model that matches a 2.5 GB teacher's accuracy and transcribes an hour of audio in 20 seconds on a MacBook Air.
A RoBERTa classifier from Unitary scores comments across seven toxicity and identity-attack labels while actively minimizing demographic bias during training.
An ambitious MiniMax H3 VAE project failed its original goal but shipped a working 2X decoder and reference detail enhancer for ComfyUI.
A community fine-tune of Alibaba's compact Qwen-Image-2.1 delivers uncensored anime illustration in an INT8 build that runs on 12GB consumer GPUs.
A 78 MB ONNX port of OpenAI's Whisper Tiny English model is quietly powering hundreds of browser-based transcription apps, no server required.
A compact ELECTRA-based cross-encoder trained on MS MARCO for reranking search results, now shipping with ONNX, OpenVINO, and Safetensors weights.
Pre-quantized GGUF weights and a portable C++ backend let YuE2 generate 48kHz stereo songs on a single consumer GPU with an editable ABC score.
Google DeepMind packaged its 740M multimodal embedding model for LiteRT, letting phones embed text, images, audio and video in one 768-dim space.
NVIDIA published a 4-bit NVFP4 build of Qwen2.5-VL-7B-Instruct that runs on Blackwell Tensor Cores through TensorRT-LLM, cutting memory roughly 3.5x.
A CTranslate2 conversion of OpenAI's Whisper medium delivers up to 4x faster transcription with lower memory, and it's crossed 830K downloads.
NVIDIA's BigVGAN v2 vocoder now runs at 44kHz with 512x upsampling, a fused CUDA kernel for 1.5-3x faster synthesis, and training across diverse audio.
Huihui released an uncensored 180B version of Qwen3.8-Flash-Next using abliteration, with GGUF quants supporting a 262K context window.
A 12B multimodal classifier fine-tuned from Gemma 4 that returns calibrated probabilities across text, image, audio and video inputs.
Supersonic Labs released Julia-1, a 144M-parameter open decision model that turns a state, question, and options into one typed choice, running in-browser on WebGPU.
OrcaRouter compressed a 27B uncensored cyber-focused Qwen model from 54.7 GB to 15.7 GB while keeping 94.4% token agreement with the original.
A domain-specific BERT family pretrained on 12GB of legislation, court cases, and contracts is hitting 833K monthly downloads on Hugging Face.