Start with your machine
Lizard considers your CPU, RAM, GPU, and available graphics memory when planning a model.
Private AI for Windows
Private AI, running on your Windows PC.
Choose models that fit your Windows PC. Run them locally, then put them to work through chat and a local API.
For Windows 10 & 11. Choose a model suited to your hardware.
Windows 10 & 11Two native enginesLocal inference after setupGGUF models
A practical route to local AI
A model plan that starts with your hardware.
- 1
Inspect your hardware
Start with your CPU, memory, and graphics hardware.
- 2
Choose a fitting model
Find a model and quantization suited to your machine.
- 3
Run it locally
Use the active model in chat or through a local API.
Built for useful local work.
From choosing a model to using its output, the operating steps stay visible.
Find a model that fits
Choose compatible GGUF models and quantizations, then verify the selected download.
Work in local chat
Keep conversations and use available document and chart capabilities in the local workspace.
Connect your applications
Use the local OpenAI-compatible endpoint with applications that support a configurable provider.
Keep inference local
After activation and setup, run installed models on the Windows PC.
Inspect the evidence
Keep provider lanes and workload conditions visible when comparing performance.
A workspace on your machine.
Local chat, with useful visual output.
Supplied product preview. Interface and available features may vary by version.
Inside Lizard
Answers you can work with. Evidence you can inspect.
Lizard’s local chat can turn a conversation into visual output. Benchmark results belong to the machine, model, and workload used to produce them. Keep those conditions attached to every comparison.
Read the benchmark methodologyTwo paths.
One hardware-aware decision.
The execution engines remain distinct in the architecture and in the benchmark evidence.
Lizard Native
A resident-GPU execution path using Direct3D 12 on supported Windows hardware. Model lifecycle and warmed shader caching stay part of the runtime.
Caterpillar
A separate execution engine using a typed graph, a reusable activation arena, and a pre-compiled compute plan.
Which computers can run Lizard?
Lizard is designed for 64-bit Windows 10 and Windows 11. Model suitability and performance depend on your hardware, available memory, model, and drivers. Review the download requirements before installing.
Does it work without an internet connection?
Installed models can perform local inference after account activation and model setup. Downloads, activation, and optional connected features may require the internet.
Do the two engines run at the same time?
Lizard Native and Caterpillar are separate execution paths. Routing depends on the model and hardware. The illustration shows the decision process, not a guarantee of simultaneous execution.
Is a model included in the purchase?
Model onboarding is a separate setup step. Review the selected model’s license and requirements. The official download page explains current activation terms and optional capabilities.
Choose what you want to do next.
A closer look.
Short films and original product interfaces. Explore at your own pace.
Inside Lizard
A silent conceptual product film from the Lizard collection.
Work with your local AI
Original product preview. Select the image to explore the details.