Last Generation Paid Per Thought

    The DX-M2 Era

    You are not renting intelligence anymore. You own it.

    Ultra-low Power
    ~ 0 W
    World’s First 2nm
    0 nm
    Frontier-class models. Off the grid
    ~ 0 B
    NPU Performance
    0 TOPS
    Memory Bandwidth
    0 GB/s
    Faster than You Read (Single-Batch)
    ~ 0 TPS
    *Specifications are subject to change without notice during development.

    Chip Line-up

    How Big Should It Think?

    Others count operations. We count how much mind fits on the die.

    DX-M2FcBGA 16x16

    External Memory
    Up to 96GB
    (LPDDR5X 24GB x 4EA)

    Power Consumption
    5W (Only SoC)
    Max Bandwidth
    8-Channel (153.6 GB/s)
    Form Factor
    M.2 Module or PCIe Card

    DX-M2MFcMCM 19x19 (Tentative)

    Internal Memory

    Up to 24GB
    (LPDDR5X 24GB x 1EA)

    Power Consumption
    5W (SoC 3W + DRAM 2W)
    Max Bandwidth
    4-Channel (76.8 GB/s)
    Form Factor
    M.2 Module

    DX-M2M ProFcMCM 29x21 (Tentative)

    Internal Memory
    Up to 48GB
    (LPDDR5X 24GB x 2EA)

    Power Consumption
    8.5W (SoC 4.5W + DRAM 4W)
    Max Bandwidth
    8-Channel (153.6 GB/s)
    Form Factor
    M.2 Module
    DX-M2 Chip Line-up Brief

    Every Deployment Breaks Even

    Here Sooner Than You Think

    The cloud bills you for thinking. The DX-M2 bills you once.
    The Cloud AI Breakeven Point
    Dependency Trend by Model Size
    Data Center Reliance (%)
    Generative AI Model Size (Billion Parameters, 20B ~ 100B)
    Shifting Workloads
    Physical AI Takes the Edge
    Deployment of Generative at the Edge

    Cloud Bottlenecks

    Infrastructure expansion is constrained by power grid volatility and skyrocketing cloud costs.

    Native Edge Demand

    The market demands native LLM execution on high-performance edge devices.

    The DX-M2 Solution

    DX-M2 resolves these constraints with architectural innovations for next-gen physical AI.
    DXNN SDK for DX-M2

    Your Code Already Runs Here.

    Nothing to port. Nothing to rewrite.
    The compiler does the work you were planning to do.
    01

    Connect

    Bring Your Own Stack

    Seamlessly run your existing frameworks and APIs on DX-M2 with zero code modification.

    • OpenAI-compatible API
    • PyTorch
    • llama.cpp
    • Ollama
    • vLLM
    • ExecuTorch
    02

    Optimize

    Tuned to the Metal

    Unlock peak hardware performance with an M2-native SDK that extracts every TOPS from silicon.
    • DX-LLM: M2-native Runtime Framework
    • Model / Memory / Compute Optimization
    • NPU / DSP Compute Library
    03

    Unify

    One Layer to Run It All

    Consolidate standard frameworks and DX-LLM workloads under one unified middleware layer.

    • M2 Middleware · Driver HAL
    • Unified Compiler & Custom Kernel Build
    Universal Model Ecosystem

    DX-M2 Optimized Model Zoo.

    Every model is pre-optimized for the DX-M2 NPU and served through DX-LLM.
    These are the models we measured ourselves — quantization variants benchmarked
    against the bf16 baseline, on-device, with zero token fees.

    Featured Models

    Text-to-Image
    Stability AI
    stable-diffusion-3.5-medium
    Accuracy
    GPU Baseline
    62.08
    BF16
    NPU Quantized
    62.08
    +0.00
    W8A8
    Text-to-Image
    Stability AI
    stable-diffusion-3.5-large
    Accuracy
    GPU Baseline
    63.14
    BF16
    NPU Quantized
    63.68
    +0.54
    W8A8
    Text Generation
    Meta
    Llama-3.2-1B
    Accuracy
    GPU Baseline
    48.14
    BF16
    NPU Quantized
    48.28
    +0.14
    W8A8
    Speech Recognition
    OpenAI
    whisper-large-v3
    Accuracy
    GPU Baseline
    89.97
    BF16
    NPU Quantized
    89.97
    +0.00
    W8A8
    Speech Recognition
    OpenAI
    whisper-large-v3-turbo
    Accuracy
    GPU Baseline
    89.09
    BF16
    NPU Quantized
    89.13
    +0.04
    W8A8
    Text Generation
    Alibaba
    Qwen3-4B
    Accuracy
    GPU Baseline
    65.97
    BF16
    NPU Quantized
    66.14
    +0.16
    W8A8
    Text Generation
    Meta
    Llama-3.1-8B
    Accuracy
    GPU Baseline
    66.70
    BF16
    NPU Quantized
    67.27
    +0.57
    W8A8
    Image-Text-to-Text
    Alibaba
    Qwen3-VL-8B
    Accuracy
    GPU Baseline
    75.25
    BF16
    NPU Quantized
    75.53
    +0.28
    W8A8
    Image-Text-to-Text
    Alibaba
    Qwen3-VL-2B
    Accuracy
    GPU Baseline
    67.36
    BF16
    NPU Quantized
    67.36
    +0.00
    W8A8

    Explore All Models

    DX-M2-btn
    Model
    Task
    Input
    Output
    Result
    Qwen3-VL-2BAlibaba
    Image-Text-to-Text
    Text
    Image
    Video
    Text
    Deploy
    Qwen3-VL-8BAlibaba
    Image-Text-to-Text
    Text
    Image
    Video
    Text
    Deploy
    gpt-oss-20bOpenAI
    Text Generation
    Text
    Text
    Deploy
    Llama-3.1-8BMeta
    Text Generation
    Text
    Text
    Deploy
    DeepSeek-R1-Distill-Llama-8BDeepSeek
    Text Generation
    Text
    Text
    Deploy
    Qwen3-4BAlibaba
    Text Generation
    Text
    Text
    Deploy
    Qwen3-1.7BAlibaba
    Text Generation
    Text
    Text
    Deploy
    Qwen3-0.6BAlibaba
    Text Generation
    Text
    Text
    Deploy
    whisper-large-v3-turboOpenAI
    Speech Recognition
    Video
    Audio
    Deploy
    whisper-large-v3OpenAI
    Speech Recognition
    Audio
    Audio
    Deploy
    gemma-4-E2BGoogle
    Image-Text-to-Text
    Text
    Image
    Video
    Audio
    Text
    Deploy
    gemma-4-E4BGoogle
    Image-Text-to-Text
    Text
    Image
    Video
    Audio
    Text
    Deploy
    Llama-3.2-1BMeta
    Text Generation
    Text
    Text
    Deploy
    stable-diffusion-3.5-largeStability AI
    Text-to-Image
    Text
    Image
    Deploy
    stable-diffusion-3.5-mediumStability AI
    Text-to-Image
    Text
    Image
    Deploy
    Phi-4-miniMicrosoft
    Text Generation
    Text
    Text
    Deploy
    FLUX.2-klein-4BBlack Forest Labs
    Text-to-Image
    Text
    Image
    Image
    Deploy
    Ministral-8B-2410Mistral AI
    Text Generation
    Text
    Text
    Deploy
    Qwen2.5-Coder-7BAlibaba
    Text Generation
    Text
    Text
    Deploy
    Qwen2.5-Coder-1.5BAlibaba
    Text Generation
    Text
    Text
    Deploy
    Qwen2.5-7BAlibaba
    Text Generation
    Text
    Text
    Deploy
    Qwen2.5-0.5BAlibaba
    Text Generation
    Text
    Text
    Deploy
    Qwen1.5-MoE-A2.7BAlibaba
    Text Generation
    Text
    Text
    Deploy
    Mistral-7B-v0.3Mistral AI
    Text Generation
    Text
    Text
    Deploy
    Phi-3.5-visionMicrosoft
    Image-Text-to-Text
    Text
    Image
    Video
    Text
    Deploy
    EXAONE-4.0-1.2BLG AI Research
    Text Generation
    Text
    Text
    Deploy
    InternVL3.5-2BOpenGVLab
    Image-Text-to-Text
    Text
    Image
    Video
    Text
    Deploy
    InternVL3.5-8BOpenGVLab
    Image-Text-to-Text
    Text
    Image
    Video
    Text
    Deploy
    GR00T-N1.7-3BNVIDIA
    Robotics
    Text
    Image
    Video
    Action
    Deploy
    Alpamayo-R1-10BNVIDIA
    Robotics
    Text
    Image
    Video
    Action
    Deploy
    segformer-b0NVIDIA
    Segmentation
    Image
    Mask
    Deploy
    kanana-2-3b-instructKakao
    Text Generation
    Text
    Text
    Deploy
    kanana-2-3b-baseKakao
    Text Generation
    Text
    Text
    Deploy
    EXAONE-3.5-32BLG AI Research
    Text Generation
    Text
    Text
    Deploy
    EXAONE-3.5-7.8BLG AI Research
    Text Generation
    Text
    Text
    Deploy
    EXAONE-3.5-2.4BLG AI Research
    Text Generation
    Text
    Text
    Deploy
    EXAONE-3.0-7.8BLG AI Research
    Text Generation
    Text
    Text
    Deploy
    openvla-7bOpenVLA
    Robotics
    Text
    Image
    Video
    Action
    Deploy
    SOLAR-10.7B-v1.0Upstage
    Text Generation
    Text
    Text
    Deploy
    pi05Physical Intelligence
    Robotics
    Text
    Image
    Video
    Action
    Deploy
    pi0Physical Intelligence
    Robotics
    Text
    Image
    Video
    Action
    Deploy
    HyperCLOVAX-SEED-Text-1.5BNAVER
    Text Generation
    Text
    Text
    Deploy
    HyperCLOVAX-SEED-Text-0.5BNAVER
    Text Generation
    Text
    Text
    Deploy
    HyperCLOVAX-SEED-Vision-3BNAVER
    Text Generation
    Text
    Text
    Deploy
    A.X-4.0-LightSK Telecom
    Text Generation
    Text
    Text
    Deploy
    A.X-3.1-LightSK Telecom
    Text Generation
    Text
    Text
    Deploy
    OLMoE-1B-7B-0924Allen AI
    Text Generation
    Text
    Text
    Deploy
    soda-4b-baseSoda Research
    Speech Recognition
    Text
    Audio
    Text
    Audio
    Deploy
    h2o-danube3-4bH2O.ai
    Text Generation
    Text
    Text
    Deploy
    mask2formerMeta
    Segmentation
    Image
    Mask
    Deploy
    yolos-baseHUST Vision
    Object Detection
    Image
    Box
    Class
    Deploy
    yolos-smallHUST Vision
    Object Detection
    Image
    Box
    Class
    Deploy
    AlexNetTorchVision
    Classification
    Image
    Class
    Deploy
    ConvNeXtTinyTorchVision
    Classification
    Image
    Class
    Deploy
    ConvNeXtSmallTorchVision
    Classification
    Image
    Class
    Deploy
    ConvNeXtBaseTorchVision
    Classification
    Image
    Class
    Deploy
    ConvNeXtLargeTorchVision
    Classification
    Image
    Class
    Deploy
    DenseNet121TorchVision
    Classification
    Image
    Class
    Deploy
    DenseNet161TorchVision
    Classification
    Image
    Class
    Deploy
    DenseNet169TorchVision
    Classification
    Image
    Class
    Deploy
    DenseNet201TorchVision
    Classification
    Image
    Class
    Deploy
    EfficientNetB0
    Classification
    Image
    Class
    Deploy
    EfficientNetB1TorchVision
    Classification
    Image
    Class
    Deploy
    EfficientNetB2TorchVision
    Classification
    Image
    Class
    Deploy
    EfficientNetB3TorchVision
    Classification
    Image
    Class
    Deploy
    EfficientNetB4TorchVision
    Classification
    Image
    Class
    Deploy
    EfficientNetB5TorchVision
    Classification
    Image
    Class
    Deploy
    EfficientNetB6TorchVision
    Classification
    Image
    Class
    Deploy
    EfficientNetB7TorchVision
    Classification
    Image
    Class
    Deploy
    EfficientNetV2STorchVision
    Classification
    Image
    Class
    Deploy
    EfficientNetV2MTorchVision
    Classification
    Image
    Class
    Deploy
    EfficientNetV2LTorchVision
    Classification
    Image
    Class
    Deploy
    GoogLeNetTorchVision
    Classification
    Image
    Class
    Deploy
    InceptionV3TorchVision
    Classification
    Image
    Class
    Deploy
    MaxVitTTorchVision
    Classification
    Image
    Class
    Deploy
    MNASNet0_5TorchVision
    Classification
    Image
    Class
    Deploy
    MNASNet1_0TorchVision
    Classification
    Image
    Class
    Deploy
    MNASNet1_3TorchVision
    Classification
    Image
    Class
    Deploy
    MobileNetV2TorchVision
    Classification
    Image
    Class
    Deploy
    MobileNetV3LargeTorchVision
    Classification
    Image
    Class
    Deploy
    MobileNetV3SmallTorchVision
    Classification
    Image
    Class
    Deploy
    RegNetY400MFTorchVision
    Classification
    Image
    Class
    Deploy
    RegNetY800MFTorchVision
    Classification
    Image
    Class
    Deploy
    RegNetY1_6GFTorchVision
    Classification
    Image
    Class
    Deploy
    RegNetY3_2GFTorchVision
    Classification
    Image
    Class
    Deploy
    RegNetY8GFTorchVision
    Classification
    Image
    Class
    Deploy
    RegNetY16GFTorchVision
    Classification
    Image
    Class
    Deploy
    RegNetY32GFTorchVision
    Classification
    Image
    Class
    Deploy
    RegNetX400MFTorchVision
    Classification
    Image
    Class
    Deploy
    RegNetX800MFTorchVision
    Classification
    Image
    Class
    Deploy
    RegNetX1_6GFTorchVision
    Classification
    Image
    Class
    Deploy
    RegNetX3_2GFTorchVision
    Classification
    Image
    Class
    Deploy
    RegNetX8GFTorchVision
    Classification
    Image
    Class
    Deploy
    RegNetX16GFTorchVision
    Classification
    Image
    Class
    Deploy
    RegNetX32GFTorchVision
    Classification
    Image
    Class
    Deploy
    ResNet18TorchVision
    Classification
    Image
    Class
    Deploy
    ResNet34TorchVision
    Classification
    Image
    Class
    Deploy
    ResNet50TorchVision
    Classification
    Image
    Class
    Deploy
    ResNet101TorchVision
    Classification
    Image
    Class
    Deploy
    ResNet152TorchVision
    Classification
    Image
    Class
    Deploy
    ResNeXt50_32x4dTorchVision
    Classification
    Image
    Class
    Deploy
    ResNeXt101_32x8dTorchVision
    Classification
    Image
    Class
    Deploy
    ResNeXt101_64x4dTorchVision
    Classification
    Image
    Class
    Deploy
    ShuffleNetV2X0_5TorchVision
    Classification
    Image
    Class
    Deploy
    ShuffleNetV2X1_0TorchVision
    Classification
    Image
    Class
    Deploy
    ShuffleNetV2X1_5TorchVision
    Classification
    Image
    Class
    Deploy
    ShuffleNetV2X2_0TorchVision
    Classification
    Image
    Class
    Deploy
    SqueezeNet1_0TorchVision
    Classification
    Image
    Class
    Deploy
    SqueezeNet1_1TorchVision
    Classification
    Image
    Class
    Deploy
    SwinTTorchVision
    Classification
    Image
    Class
    Deploy
    SwinSTorchVision
    Classification
    Image
    Class
    Deploy
    SwinBTorchVision
    Classification
    Image
    Class
    Deploy
    SwinV2TTorchVision
    Classification
    Image
    Class
    Deploy
    SwinV2STorchVision
    Classification
    Image
    Class
    Deploy
    SwinV2BTorchVision
    Classification
    Image
    Class
    Deploy
    VGG11TorchVision
    Classification
    Image
    Class
    Deploy
    VGG11BNTorchVision
    Classification
    Image
    Class
    Deploy
    VGG13BNTorchVision
    Classification
    Image
    Class
    Deploy
    VGG16TorchVision
    Classification
    Image
    Class
    Deploy
    VGG16BNTorchVision
    Classification
    Image
    Class
    Deploy
    VGG19TorchVision
    Classification
    Image
    Class
    Deploy
    VGG19BNTorchVision
    Classification
    Image
    Class
    Deploy
    ViTB16TorchVision
    Classification
    Image
    Class
    Deploy
    ViTB32TorchVision
    Classification
    Image
    Class
    Deploy
    ViTL16TorchVision
    Classification
    Image
    Class
    Deploy
    ViTL32TorchVision
    Classification
    Image
    Class
    Deploy
    ViTH14TorchVision
    Classification
    Image
    Class
    Deploy
    WideResNet50_2TorchVision
    Classification
    Image
    Class
    Deploy
    WideResNet101_2TorchVision
    Classification
    Image
    Class
    Deploy