xenovatech

• In-browser tool calling playground, running LFM2 locally on WebGPU with Transformers.js

  Posted by xenovatech@reddit | LocalLLaMA | View on Reddit | 2 comments

• PrismML just released Binary and Ternary Bonsai Image 4B: 1-bit/ternary text-to-image diffusion transformers that can even run 100% locally in your browser on WebGPU.

  Posted by xenovatech@reddit | LocalLLaMA | View on Reddit | 79 comments

• 1-bit Bonsai 1.7B (290MB in size) running locally in your browser on WebGPU

  Posted by xenovatech@reddit | LocalLLaMA | View on Reddit | 171 comments

• Gemma 4 running fully offline on WebGPU with Transformers.js, controlling Reachy Mini over WebSerial.

  Posted by xenovatech@reddit | LocalLLaMA | View on Reddit | 10 comments

• New OpenAI Privacy Filter model, running locally in your browser on WebGPU

  Posted by xenovatech@reddit | LocalLLaMA | View on Reddit | 0 comments

• Gemma 4 WebGPU: Run Google's new open model locally in your browser

  Posted by xenovatech@reddit | LocalLLaMA | View on Reddit | 2 comments

• Kokoro WebGPU: Real-time text-to-speech running 100% locally in your browser.

  Posted by xenovatech@reddit | LocalLLaMA | View on Reddit | 103 comments

• SAM 3: Segment Anything with Concepts, by Meta Superintelligence Labs

  Posted by xenovatech@reddit | LocalLLaMA | View on Reddit | 22 comments

• Cohere Transcribe WebGPU: state-of-the-art multilingual speech recognition in your browser

  Posted by xenovatech@reddit | LocalLLaMA | View on Reddit | 1 comments

• Liquid AI's LFM2-24B-A2B running at ~50 tokens/second in a web browser on WebGPU

  Posted by xenovatech@reddit | LocalLLaMA | View on Reddit | 18 comments

• Nemotron-3-Nano (4B), new hybrid Mamba + Attention model from NVIDIA, running locally in your browser on WebGPU.

  Posted by xenovatech@reddit | LocalLLaMA | View on Reddit | 6 comments

• Running Qwen 3.5 0.8B locally in the browser on WebGPU w/ Transformers.js

  Posted by xenovatech@reddit | LocalLLaMA | View on Reddit | 32 comments

• Real-time video captioning in the browser with LFM2-VL on WebGPU

  Posted by xenovatech@reddit | LocalLLaMA | View on Reddit | 2 comments

• Voxtral WebGPU: Real-time speech transcription entirely in your browser with Transformers.js

  Posted by xenovatech@reddit | LocalLLaMA | View on Reddit | 13 comments

• microgpt playground: Build, train, and run LLMs — directly in your browser

  Posted by xenovatech@reddit | LocalLLaMA | View on Reddit | 10 comments

• Run LFM2.5-1.2B-Thinking at over 200 tokens per second in your browser on WebGPU

  Posted by xenovatech@reddit | LocalLLaMA | View on Reddit | 11 comments

• Text Behind Video: Create cinematic text and video compositions locally in your browser w/ Transformers.js

  Posted by xenovatech@reddit | LocalLLaMA | View on Reddit | 0 comments

• GPT-OSS (20B) running 100% locally in your browser on WebGPU

  Posted by xenovatech@reddit | LocalLLaMA | View on Reddit | 26 comments

• Supertonic WebGPU: blazingly fast text-to-speech running 100% locally in your browser.

  Posted by xenovatech@reddit | LocalLLaMA | View on Reddit | 11 comments

• Real-time conversational AI running 100% locally in-browser on WebGPU

  Posted by xenovatech@reddit | LocalLLaMA | View on Reddit | 135 comments

• Chatterbox Turbo, new open-source voice AI model, just released on Hugging Face

  Posted by xenovatech@reddit | LocalLLaMA | View on Reddit | 64 comments

• FunctionGemma Physics Playground: A simulation game where you need to use natural language to solve physics puzzles... running 100% locally in your browser!

  Posted by xenovatech@reddit | LocalLLaMA | View on Reddit | 17 comments

• Ministral WebGPU: Run Mistral's new multimodal models 100% locally in your browser.

  Posted by xenovatech@reddit | LocalLLaMA | View on Reddit | 15 comments

• Run Qwen3 (0.6B) 100% locally in your browser on WebGPU w/ Transformers.js

  Posted by xenovatech@reddit | LocalLLaMA | View on Reddit | 21 comments

• IBM releases Granite-4.0 Nano (300M & 1B), along with a local browser demo showing how the models can programmatically interact with websites and call tools/browser APIs on your behalf.

  Posted by xenovatech@reddit | LocalLLaMA | View on Reddit | 42 comments

• NanoChat WebGPU: Karpathy's full-stack ChatGPT project running 100% locally in the browser.

  Posted by xenovatech@reddit | LocalLLaMA | View on Reddit | 5 comments

• Granite Docling WebGPU: State-of-the-art document parsing 100% locally in your browser.

  Posted by xenovatech@reddit | LocalLLaMA | View on Reddit | 46 comments

• Granite 4.0 Micro (3.4B) running 100% locally in your browser w/ WebGPU acceleration

  Posted by xenovatech@reddit | LocalLLaMA | View on Reddit | 47 comments

• In-browser video object detection w/ YOLOv9 and Transformers.js

  Posted by xenovatech@reddit | LocalLLaMA | View on Reddit | 2 comments

• Apple releases FastVLM and MobileCLIP2 on Hugging Face, along with a real-time video captioning demo (in-browser + WebGPU)

  Posted by xenovatech@reddit | LocalLLaMA | View on Reddit | 162 comments

• DINOv3 semantic video tracking running locally in your browser (WebGPU)

  Posted by xenovatech@reddit | LocalLLaMA | View on Reddit | 19 comments

• The Semantic Galaxy: An interactive 3D embedding visualization demo, built with Google's new EmbeddingGemma model

  Posted by xenovatech@reddit | LocalLLaMA | View on Reddit | 11 comments

• DINOv3 visualization tool running 100% locally in your browser on WebGPU/WASM

  Posted by xenovatech@reddit | LocalLLaMA | View on Reddit | 34 comments

• In-browser background removal w/ RMBG-v1.4 and Transformers.js

  Posted by xenovatech@reddit | LocalLLaMA | View on Reddit | 2 comments

• Voxtral WebGPU: State-of-the-art audio transcription directly in your browser!

  Posted by xenovatech@reddit | LocalLLaMA | View on Reddit | 13 comments

• Introducing Kokoro.js: a new JavaScript library for running Kokoro TTS (82M) locally in the browser w/ WASM.

  Posted by xenovatech@reddit | LocalLLaMA | View on Reddit | 60 comments

• WebGPU-accelerated real-time in-browser speech recognition w/ Transformers.js

  Posted by xenovatech@reddit | LocalLLaMA | View on Reddit | 63 comments

• I updated the SmolVLM llama.cpp webcam demo to run locally in-browser on WebGPU.

  Posted by xenovatech@reddit | LocalLLaMA | View on Reddit | 27 comments

• Whisper Timestamped: Multilingual speech recognition w/ word-level timestamps, running locally in your browser using Transformers.js

  Posted by xenovatech@reddit | LocalLLaMA | View on Reddit | 54 comments

• Hugging Face Chat Templates, now available in Transformers.js

  Posted by xenovatech@reddit | LocalLLaMA | View on Reddit | 2 comments

• Apply formatting to Jinja chat templates directly from the Hugging Face model card (+ new playground)

  Posted by xenovatech@reddit | LocalLLaMA | View on Reddit | 4 comments

• ONNX Model Explorer and Visualization Tool

  Posted by xenovatech@reddit | LocalLLaMA | View on Reddit | 0 comments

• Whisper Diarization Web: In-browser multilingual speech recognition with word-level timestamps and speaker segmentation

  Posted by xenovatech@reddit | LocalLLaMA | View on Reddit | 32 comments

• Introducing Kokoro Web: ML-powered speech synthesis directly in your browser. Now with streaming & WebGPU acceleration.

  Posted by xenovatech@reddit | LocalLLaMA | View on Reddit | 19 comments

• Running Llama 3.2 100% locally in the browser on WebGPU w/ Transformers.js

  Posted by xenovatech@reddit | LocalLLaMA | View on Reddit | 38 comments

• Janus, a new multimodal understanding and generation model from Deepseek, running 100% locally in the browser on WebGPU with Transformers.js!

  Posted by xenovatech@reddit | LocalLLaMA | View on Reddit | 25 comments

• Transformers.js v3 is finally out: WebGPU Support, New Models & Tasks, New Quantizations, Deno & Bun Compatibility, and More…

  Posted by xenovatech@reddit | LocalLLaMA | View on Reddit | 26 comments

• Janus Pro 1B running 100% locally in-browser on WebGPU, powered by Transformers.js

  Posted by xenovatech@reddit | LocalLLaMA | View on Reddit | 70 comments

• DeepSeek-R1-Distill-Qwen-1.5B running 100% locally in-browser on WebGPU. Reportedly outperforms GPT-4o and Claude-3.5-Sonnet on math benchmarks (28.9% on AIME and 83.9% on MATH).

  Posted by xenovatech@reddit | LocalLLaMA | View on Reddit | 49 comments

• SmolVLM 256M: The world's smallest multimodal model, running 100% locally in-browser on WebGPU.

  Posted by xenovatech@reddit | LocalLLaMA | View on Reddit | 13 comments