Product 01 / Lizard LLM
Lizard runs AI models locally, measures their real performance on your hardware, and provides an OpenAI-compatible endpoint for your applications.
Qendryx product system
Qendryx builds tools around the work a machine should do well: run private models locally, and turn spoken information into structured business output.
A hardware-aware local AI runtime that plans around the CPU, memory, and graphics hardware already inside a Windows PC. It recommends models that fit, verifies the download, and routes inference through the right native engine.
View the Lizard product profileCapture meetings, browser audio, and files; then create transcripts, summaries, action items, and workflow-ready business reports.
Open SonicMind ConnectNative execution architecture
Lizard selects the execution path for the model and hardware. The distinction remains visible, measurable, and documented.
The stable Direct3D 12 path for supported Windows GPUs. It keeps model weights resident across calls and uses the machine’s graphics hardware for local GGUF inference.
A second standalone engine built around a pre-compiled compute-graph plan. It is designed to reduce repeated orchestration work and provide another path across CPU and shared or integrated graphics systems.
System route / simplified
From hardware to answer
Local AI should begin with the machine, not a guess. Lizard turns hardware facts into a usable model plan and keeps the operating steps visible.
Lizard reads the CPU, system memory, GPU, and available graphics memory. The resulting profile becomes the basis for every recommendation.
The app recommends compatible GGUF models and quantizations, then downloads and verifies the selected file before activation.
Use the built-in chat workspace or connect existing software through the local OpenAI-compatible API. Installed models can run without an internet connection.
Inspectable by design
Local performance depends on the exact hardware, model, quantization, driver, and workload. Lizard’s benchmark workspace keeps provider lanes separate and records the conditions behind each result.
Product 02 / SonicMind Connect
SonicMind Connect turns meetings, browser audio, voice notes, and uploaded files into transcripts, summaries, action items, follow-up questions, and business-ready reports.
It remains the connected web product in the Qendryx portfolio—separate from Lizard’s local runtime, and available through its existing application and specialist pages.
Qendryx design principles
Different products, consistent discipline: start with the environment, keep the workflow understandable, and give users a direct route to useful output.
Lizard begins with the hardware already present. SonicMind begins with the sources and workflows a team already uses.
Model plans, engine selection, benchmark evidence, report structure, and follow-up work remain part of the product interface.
A local model should be callable. A conversation should become work. Qendryx products are designed around the completed outcome.
Questions
Lizard is a hardware-aware local AI runtime for 64-bit Windows 10 and Windows 11. It profiles the machine, recommends compatible GGUF models, and runs inference locally through Lizard Native, Caterpillar, or an explicit compatibility fallback.
Lizard Native is the resident Direct3D 12 GPU path. Caterpillar is a separate engine built around a pre-compiled compute-graph plan. Lizard selects the path for the current hardware and model.
Internet access is needed for sign-in, activation, model downloads, and update checks. Once a compatible model is installed, prompts, answers, and model weights used for inference remain on the device and the model can run offline.
The native Lizard runner is activated with one USD 5 payment. It is not a subscription. Optional dashboard credits are separate and do not control the local runner.
SonicMind Connect is Qendryx’s second product: a connected web workspace for AI transcription, summaries, action items, follow-up questions, and business reports. Its existing application remains available at sonicmindconnect.qendryx.com.
Lizard LLM / Windows
Install Lizard, inspect the hardware, and choose a local model plan based on what the computer can actually run.
