Search Results for "latency" - Page 27

UltraSonic is a desktop "Music Production Platform" extensible by plug-ins. Instruments and effects are edited and played back in realtime with low latency. Mastering and rendering to disk supports high quality audio results.

Open Source project Powered by WebRTC using Google Stun and numb Turn. MiroTalk provides video quality and latency not available with traditional technology. A good Free alternative to Zoom, Google-Meet, Microsoft Teams... Simple, Secure, Fast and Self-Hosted. More details here: https://github.com/miroslavpejic85/mirotalk

xpsLightFX daemon allows the vanity LED lights on select Dell XPS laptops to be controlled through DBus for minimal latency. The primary goal is to allow the LEDs to dance to music. A libvisual plugin suitable for use with Amarok is provided to this end.

Rampart is a lightweight, on-device privacy protection model developed by the National Design Studio to detect and redact personally identifiable information (PII) before text leaves a user's device. Rather than relying on server-side filtering, Rampart performs token-level PII detection locally, enabling privacy-preserving AI interactions with minimal latency and without exposing sensitive information to external services. The released model is a 14.7 MB ONNX artifact based on a fine-tuned MiniLM-L6-H384 encoder with approximately 18.5 million parameters and a 35-label BIO classification head covering 17 entity types. It works alongside a deterministic recognizer that handles structured identifiers such as phone numbers and IDs, forming a defense-in-depth client-side redaction system. ...

...It uses 125B language-model parameters with only 6B activated per token, supplemented by 51B n-gram embedding parameters and 4B for multi-token prediction. Its hybrid architecture combines Gated DeltaNet with Qwen Sparse Attention (QSA), which processes micro-blocks rather than individual tokens to reduce latency in long-context agent workloads. The model also introduces Gated Residual connections and scalable n-gram embeddings to improve efficiency while limiting inference overhead. It contains 512 MoE experts, activating 10 routed experts plus one shared expert per token. Qwen3.8-Flash-Next natively handles text, images, and video and supports a 262K-token context window extensible to 1M tokens. ...

...The model natively processes text, images, audio, and video within a unified architecture and supports an exceptionally large 1 million token context window for long-document reasoning, repository-scale coding, and agentic execution. Trained from scratch on approximately 45 trillion multimodal tokens, Inkling introduces controllable reasoning effort, allowing users to trade off latency and reasoning depth depending on the task. It is optimized for software engineering, tool use, and large-scale autonomous workflows, with strong performance on coding and agent benchmarks.

Nex-N2-mini is an open-source agentic model from Nex AGI designed for real-world productivity, coding, tool use, deep research, and terminal-based execution. Built on Qwen3.5-35B-A3B-Base, it offers a lighter latency and deployment profile than Nex-N2-Pro while preserving the core Nex-N2 “Agentic Thinking” framework. This framework unifies requirement understanding, planning, code implementation, environmental feedback, debugging, evaluation, and iteration into a closed loop. It uses adaptive thinking to decide when deeper reasoning is needed and coherent thinking to keep reasoning consistent across tasks and modalities. ...

...Unlike other Gemma 4 models that rely on separate encoders, the 12B Unified model uses an encoder-free architecture that projects raw image patches and audio waveforms directly into the language model’s embedding space, reducing multimodal latency and simplifying fine-tuning. It supports text, image, audio, and video inputs with text output, making it useful for transcription, image understanding, video analysis, coding, and agentic workflows. The model has 11.95B parameters, 48 layers, a 256K-token context window, and support for over 140 languages. It also includes configurable thinking modes, native system prompt support, function calling, and strong benchmark performance for its size. ...

...The W4A4 release applies 4-bit weight and activation quantization mainly to MoE experts, preserving attention components at full precision to reduce quality loss while improving speed, latency, and hardware efficiency. Cohere recommends W4A4 for most users because it offers a smaller hardware footprint with negligible benchmark differences compared to BF16 and FP8 versions. The model supports a 128K input context and 64K output length, covers 48 languages, and includes conversational tool-use capabilities with JSON-schema tools and optional citation grounding.

...The model supports eight primary languages (including English, Spanish, Hindi, and Thai) and was trained on a curated mix of publicly available online data, with a December 2023 knowledge cutoff. Llama-3.2-1B is lightweight enough for deployment on constrained devices like smartphones, using formats like SpinQuant and QLoRA to reduce model size and latency. Despite its small size, it performs competitively across benchmarks such as MMLU, ARC, and TLDR summarization. The model is distributed under the Llama 3.2 Community License, requiring attribution and adherence to Meta’s Acceptable Use Policy.

Monitoring tool with support to Websites, RSS, Webservices and Databases. Has notifications by email and RSS and you can access metrics like availability, latency and load time by a web-based GUI. Runs standalone with an embedded HTTP server.