Browse supported models by task, hardware and release status.
The counts are the catalogue and release registry as built. A released family can list variants that are not in a release yet; each card says how many are.
Run Qwen2.5 directly from supported safetensors checkpoints with no converted model copy.
Run the post-trained Qwen3-1.7B checkpoint in a compact native runtime.
Qwen3.6-35B-A3B mixture-of-experts; on NVIDIA it also takes an image or a short video.
Text-to-video and image-to-video with Wan2.1 and Wan2.2 checkpoints.
Still-image generation from text, run directly from the published checkpoints.
Still-image generation from text, read straight from the published safetensors checkpoints.
Run Stable Diffusion 1.5 in Veizik's native edge runtime on a Raspberry Pi CPU — no GPU, no Python.
Run SDXS in Veizik's native edge runtime on a Raspberry Pi CPU — no GPU, no Python.
Run SmolLM2 in Veizik's native edge runtime on a Raspberry Pi CPU — no GPU, no Python.
Run GPT-2 in Veizik's native edge runtime on a Raspberry Pi CPU — no GPU, no Python.
Run LFM2.5 in Veizik's native edge runtime on a Raspberry Pi CPU — no GPU, no Python.
Short text-to-video on a Raspberry Pi 5 CPU: AnimateDiff-Lightning motion on a DreamShaper-8 base.
Soft Actor-Critic policy for Pendulum-v1, run natively on an Exynos / Android SoC.
TD3 policy for BipedalWalker-v3, run natively on an Exynos / Android SoC.
Soft Actor-Critic policy for Humanoid-v3, run natively on an Exynos / Android SoC.
TD3 policy for HalfCheetah-v3, run natively on an Exynos / Android SoC.
Qwen3.5 language models, from 0.8B dense checkpoints to a 122B mixture-of-experts flagship.
Still-image generation from text, read from the published checkpoints.
Image-generation checkpoints in a format the engine supports. Not offered as a download.
Video-generation path in the engine. Not offered as a download.
Run NVIDIA's SANA diffusion models in Veizik's native edge runtime on a Raspberry Pi CPU.
Run OpenAI's gpt-oss open-weight mixture-of-experts models in Veizik's native runtime.
Soft Actor-Critic whole-body humanoid control as a real-time loop on a Jetson AGX Orin.
Soft Actor-Critic quadruped locomotion control as a real-time loop on a Jetson AGX Orin.
TD3 bipedal locomotion control as a real-time loop on a Jetson AGX Orin.
TD3 running-gait locomotion control as a real-time loop on a Jetson AGX Orin.
| Platforms | Build | Measured memory | ||||
|---|---|---|---|---|---|---|
| Qwen2.5Alibaba (Qwen) | Language | Available | Apple Silicon | 2026.10.02-r10Support rows measured on 2026.10.01 | 6 | 344–4,114 MBWeights in memory |
| Qwen3Alibaba (Qwen) | Language | Available | Apple Silicon | 2026.10.02-r10Support rows measured on 2026.10.01 | 2 | 968 MBWeights in memory |
| Qwen3.6Alibaba (Qwen) | Language | Available | NVIDIA GPU | 1.1.4 | 1 | 21.4 GiBGPU memory in use |
| Wan 2.1 · Wan 2.2Wan-AI (Alibaba) | Video | Available | NVIDIA GPU | 1.2.93 | 4 | Not measured |
| FLUX.1Black Forest Labs | Image | Available | NVIDIA GPU | 1.0.5 | 8 | Not measured |
| Stable Diffusion 3.5Stability AI | Image | Available | NVIDIA GPU | 1.0.1 | 2 | Not measured |
| Stable Diffusion 1.5Stability AI · RunwayML | Image | Available | Raspberry Pi | — | 1 | Not measured |
| SDXSIDKiro (SDXS) | Image | Available | Raspberry Pi | — | 1 | Not measured |
| SmolLM2Hugging Face | Language | Available | Raspberry Pi | — | 2 | Not measured |
| GPT-2OpenAI | Language | Available | Raspberry Pi | — | 1 | Not measured |
| LFM2.5Liquid AI | Language | Available | Raspberry Pi | — | 1 | Not measured |
| AnimateDiff-LightningByteDance (AnimateDiff-Lightning) · Lykon (DreamShaper-8) | Video | Available | Raspberry Pi | — | 1 | Not measured |
| SAC · PendulumStable-Baselines3 RL Zoo | Robotics | Available | Android · Exynos | — | 1 | Not measured |
| TD3 · BipedalWalkerStable-Baselines3 RL Zoo | Robotics | Available | Android · Exynos | — | 1 | Not measured |
| SAC · HumanoidStable-Baselines3 RL Zoo | Robotics | Available | Android · Exynos | — | 1 | Not measured |
| TD3 · HalfCheetahStable-Baselines3 RL Zoo | Robotics | Available | Android · Exynos | — | 1 | Not measured |
| Qwen3.5Alibaba (Qwen) | Language | Available4 of 5 variants released | NVIDIA GPU | 1.1.4 | 5 | 2.3–8.4 GiBGPU memory in use |
| FLUX.2Black Forest Labs | Image | Compatible | Not announced | — | 1 | Not measured |
| HunyuanImageTencent | Image | Compatible | Not announced | — | 2 | Not measured |
| HunyuanVideoTencent | Video | Compatible | Not announced | — | 2 | Not measured |
| SANANVIDIA | Image | In development | Raspberry Pi | — | 4 | Not measured |
| gpt-ossOpenAI | Language | In development | Not announced | — | 2 | Not measured |
| SAC · Humanoid (Jetson)Stable-Baselines3 RL Zoo | Robotics | In development | NVIDIA Jetson | — | 1 | Not measured |
| SAC · Ant (Jetson)Stable-Baselines3 RL Zoo | Robotics | In development | NVIDIA Jetson | — | 1 | Not measured |
| TD3 · BipedalWalker (Jetson)Stable-Baselines3 RL Zoo | Robotics | In development | NVIDIA Jetson | — | 1 | Not measured |
| TD3 · HalfCheetah (Jetson)Stable-Baselines3 RL Zoo | Robotics | In development | NVIDIA Jetson | — | 1 | Not measured |
Nothing is substituted. Models that are not released stay in their own tab, with their own status.
A build marked Experimental release is public and installable; its options and defaults may still change.
Weights are supplied by you from their publishers; Veizik ships the engine. To start from a device instead of a model, see hardware; to get the runtime, see download.
Stable Diffusion 3.5: Powered by Stability AI HunyuanImage: Weights are not included. Model availability and use are governed by the model provider's license and regional restrictions.