AI PORTAL
  • Models
  • Code
  • Tools

About the models on this Arm AI Portal

The Arm AI Portal (the “Portal”) provides details of AI models developed, optimized or tested for use on Arm-based platforms. Models listed on the Portal may have been developed by Arm or by third parties. Unless expressly stated otherwise, Arm did not develop or train the underlying third-party model and is not responsible for its original intended behavior or use.

Information accompanies each model listing and describes the model’s provenance, together with relevant information about the original third-party model, its developer and repository, where applicable. Inclusion on the Portal of a model developed by a third party or based on a third-party model does not constitute Arm’s endorsement or certification of that third-party model or any third-party work relating to it.

Except as expressly stated in the information accompanying a specific model, Arm makes no representation as to the accuracy, safety, security, non-infringement, legal compliance, suitability for production use or fitness for any particular purpose of any model, converted or optimized version, or related outputs.

Inclusion of a model on the Portal does not itself grant you any rights to use that model. Use of each model is subject to the applicable license terms, usage restrictions and documentation. You are responsible for reviewing that information and independently evaluating the model, its outputs and its suitability for your intended use, including compliance with applicable legal, regulatory, safety and security requirements. Any use of or reliance on a model, its outputs or related materials is at your own risk. Where the Portal links to repositories or other materials controlled by third parties, those repositories and materials are outside Arm’s control and may change without notice.

Use of the Portal is subject to the Arm Website Terms and Conditions of Use.

Models

Filters

1

Device Class

Mobile CPU

Evaluation Target

vivo X300Raspberry Pi 5AWS Graviton G4Alif DK-E8

Arm Technology

SME2

Task

Text GenerationImage ClassificationFeature ExtractionAutomatic Speech RecognitionObject DetectionImage-to-ImageText-to-SpeechDepth EstimationImage SegmentationImage-Text-to-TextText-to-ImageTranslation

Vendor

Runtime

Quantization

Format

Execution Backend

Model Size

≤5.3 GB

Average Memory

≤3.52 GB

Peak Memory

≤5.67 GB
    64 results

    YOLOX-S INT8

    vivo X300

    An INT8-quantized YOLOX-S single-stage, anchor-free object detection model for locating and classifying objects in images. Exported as an ExecuTorch program using the XNNPACK backend for inference on Arm CPUs.

    object-detection9.7 MBExecuTorch

    Whisper Large-v3 INT8

    vivo X300SME2

    INT8-quantized automatic speech recognition model for English speech-to-text, based on an encoder-decoder Transformer architecture with autoregressive decoding and KV caching. It uses 128-bin log-mel audio features and produces text transcriptions via greedy decoding, with weight-only quantization designed to reduce model size while preserving transcription quality.

    automatic-speech-recognition1.6 GBLiteRT

    Llama-3.2-1B-Base INT4

    vivo X300SME2

    A compact decoder-only causal language model for text generation, using INT4 weight quantization with dynamic INT8 activations and INT8 embeddings to reduce size and memory use while preserving practical language capabilities. It supports raw prompt continuation and in-context few-shot tasks without chat formatting.

    text-generation1.2 GBONNX

    BGE Base EN v1.5 INT8

    vivo X300

    An INT8-quantized BGE Base EN v1.5 English text embedding model for semantic search, retrieval, and similarity tasks. Exported as an ExecuTorch program using the XNNPACK backend for inference on Arm CPUs.

    feature-extraction191.8 MBExecuTorch

    Hy MT2 1.8B Q4_0

    vivo X300SME2

    Hy-MT2-1.8B machine translation optimized as a Q4_0 GGUF model for llama.cpp on Arm-based Premium Smartphone systems.

    translation1.1 GBllama.cpp

    Gemma-3-1B-Base INT4

    vivo X300SME2

    A compact decoder-only causal language model for text generation, using INT4 quantized linear weights with dynamic INT8 activations and tied INT8 block-quantized embeddings/LM head. It is designed for efficient continuation and few-shot prompting with a large SentencePiece vocabulary, offering reduced memory footprint and faster CPU inference while retaining the behavior of a base, non-instruction-tuned model.

    text-generation865.5 MBONNX

    Wide ResNet-50 INT8

    vivo X300

    An INT8-quantized Wide ResNet-50 image classification model using a widened residual convolutional network for visual feature extraction. Exported as an ExecuTorch program using the XNNPACK backend for inference on Arm CPUs.

    image-classification50.6 MBExecuTorch

    Qwen3.5-0.8B Q4_K_M

    vivo X300SME2

    This is a 4-bit weight-quantized version of Qwen/Qwen3.5-0.8B packaged as a single GGUF file for edge deployment on ARM CPUs using llama.cpp. The model was quantized with llama.cpp's native Q4_K_M k-quant recipe, guided by an importance matrix (llama-imatrix) computed over a plain-text calibration corpus, and it runs on device through ggml CPU_REPACK kernels compiled with -march=armv8.6-a+dotprod+i8mm. Norms and the SSM convolution/time-step parameters are kept in FP32. The upstream checkpoint is a vision-language model; the export produced a separate f16 multimodal projector (mmproj) GGUF that is not part of this bundle and was neither benchmarked nor evaluated. The tokenizer, merges, special tokens and chat template are embedded in the GGUF, so a single file is all you need to run the language model.

    text-generation529.3 MBllama.cpp

    Phi-4 Mini INT8

    vivo X300

    An INT8-quantized Phi-4 Mini language model for text generation. Exported as an ExecuTorch program using the XNNPACK backend for inference on Arm CPUs.

    text-generation5.3 GBExecuTorch

    TinyLlama 1.1B Chat INT8

    vivo X300

    An INT8-quantized TinyLlama 1.1B Chat causal language model for conversational text generation. Exported as an ExecuTorch program using the XNNPACK backend for inference on Arm CPUs.

    text-generation1.2 GBExecuTorch

    SmolLM2-360M-Instruct 8da4w

    vivo X300SME2

    A compact 362M-parameter decoder-only causal language model for instruction-following, open-ended text generation, and multiple-choice scoring, using 8-bit dynamic activations and 4-bit grouped weights to reduce size and memory needs while preserving core language capabilities.

    text-generation251.7 MBExecuTorch

    Whisper Base INT8

    vivo X300SME2

    Compact automatic speech recognition family for English speech-to-text, using an INT8 weight-only quantized encoder-decoder Transformer with LiteRT/TFLite support. It is designed for efficient on-device inference on ARM smartphones, with autoregressive decoding accelerated by explicit self- and cross-attention KV caching, supporting 16 kHz mono audio converted to Whisper-style log-mel features and greedy text generation.

    automatic-speech-recognition77 MBLiteRT

    Inception v3 INT8

    vivo X300

    An INT8-quantized Inception v3 convolutional image classification model using factorized convolutions for visual feature extraction. Exported as an ExecuTorch program using the XNNPACK backend for inference on Arm CPUs.

    image-classification25.6 MBExecuTorch

    Qwen2.5 0.5B Instruct INT8

    vivo X300

    An INT8-quantized Qwen2.5 0.5B Instruct causal language model for instruction-following and text generation. Exported as an ExecuTorch program using the XNNPACK backend for inference on Arm CPUs.

    text-generation806.4 MBExecuTorch

    all-MiniLM-L6-v2 INT8

    vivo X300SME2

    An INT8-quantized sentence embedding model that maps text into 384-dimensional L2-normalized vectors for semantic similarity, retrieval, and feature extraction. It uses WordPiece tokenization with mean pooling and supports inputs up to 128 tokens, offering a compact, efficient representation while preserving high embedding fidelity.

    feature-extraction23.5 MBLiteRT

    Swin-T INT8

    vivo X300

    An INT8-quantized Swin-T hierarchical Vision Transformer for image classification, using shifted-window attention for visual feature extraction. Exported as an ExecuTorch program using the XNNPACK backend for inference on Arm CPUs.

    image-classification32.5 MBExecuTorch

    Gemma 3 1B Instruct INT8

    vivo X300

    An INT8-quantized Gemma 3 1B Instruct causal language model for instruction-following and text generation. Exported as an ExecuTorch program using the XNNPACK backend for inference on Arm CPUs.

    text-generation1.7 GBExecuTorch

    MiniLM-L6-v2 INT8

    vivo X300

    An INT8-quantized MiniLM-L6-v2 sentence embedding model for semantic search, retrieval, and similarity tasks. Exported as an ExecuTorch program using the XNNPACK backend for inference on Arm CPUs.

    feature-extraction36.9 MBExecuTorch

    MobileNetV3-Small INT8

    vivo X300SME2

    Lightweight INT8 image classifier for 224×224 RGB inputs, producing logits for 1,000 ImageNet-style classes. Built for mobile and edge use on ARM CPUs, it combines a compact inverted-residual CNN architecture with quantized inference to reduce size, latency, and memory while preserving strong classification accuracy.

    image-classification3 MBLiteRT

    SmolLM3 INT8

    vivo X300

    An INT8-quantized SmolLM3 language model for text generation. Exported as an ExecuTorch program using the XNNPACK backend for inference on Arm CPUs.

    text-generation3.8 GBExecuTorch

    DeepSeek-R1-Distill-Qwen-1.5B Q4_K_M

    vivo X300SME2

    DeepSeek-R1-Distill-Qwen-1.5B text generation optimized as a Q4_K_M GGUF model for the llama.cpp runtime on Arm-based Premium Smartphone systems.

    text-generation1.1 GBllama.cpp

    Conformer CTC INT8

    vivo X300

    An INT8-quantized Conformer CTC speech recognition model for converting audio into text. Exported as an ExecuTorch program using the XNNPACK backend for inference on Arm CPUs.

    automatic-speech-recognition1.2 GBExecuTorch

    Nomic Embed Text v1 INT8

    vivo X300

    An INT8-quantized Nomic Embed Text v1 text embedding model for semantic search, retrieval, and similarity tasks. Exported as an ExecuTorch program using the XNNPACK backend for inference on Arm CPUs.

    feature-extraction168.9 MBExecuTorch

    Whisper Small INT8

    vivo X300SME2

    A compact INT8-quantized speech recognition model for transcribing 16 kHz audio, using an encoder-decoder Transformer that converts log-mel spectrograms into text tokens. It combines dynamic INT8 linear layers with selective FP32 components to reduce size and improve inference efficiency while preserving transcription quality.

    automatic-speech-recognition414.2 MBExecuTorch

    Qwen3-TTS 0.6B Base — Compute Flow

    vivo X300

    Qwen3-TTS 0.6B Base is a multilingual, controllable, robust, and streaming text-to-speech model based on the 12 Hz tokenizer. This Arm®-optimized package supports voice cloning across all original languages and produces streaming audio at 24 kHz.

    text-to-speech3.6 GBLiteRT

    DeiT-Tiny INT8

    vivo X300SME2

    This is an INT8-quantized version of facebook/deit-tiny-patch16-224 optimized for edge deployment on ARM devices using LiteRT with the XNNPACK backend and KleidiAI acceleration. The model was quantized using post-training static INT8 quantization (PT2E), calibrated on 100 ImageNet-1k images, and exported to the .tflite format for efficient inference on ARM Cortex-A processors (premium smartphone).

    image-classification7.2 MBLiteRT

    Qwen3-1.7B-Base Q4_K_M

    vivo X300SME2

    This is a 4-bit weight-quantized version of Qwen/Qwen3-1.7B-Base packaged as a single GGUF file for edge deployment on ARM CPUs using llama.cpp. The model was quantized with llama.cpp's native Q4_K_M k-quant recipe, guided by an importance matrix (llama-imatrix) computed over a plain-text calibration corpus, and it runs on device through ggml CPU_REPACK kernels compiled with -march=armv8.6-a+dotprod+i8mm. Quantization is weight-only — activations stay float at runtime, so there is no activation calibration and no runtime quantization step to configure. The tokenizer, merges, special tokens and chat template are embedded in the GGUF, so a single file is all you need to run the model.

    text-generation1.2 GBllama.cpp

    Swin-Tiny INT8

    vivo X300SME2

    An INT8-quantized hierarchical vision transformer for ImageNet-style image classification, using shifted-window self-attention to capture local and multi-scale visual features efficiently. It accepts 224×224 RGB inputs and produces logits for 1,000 classes, with reduced model size and memory needs compared with the FP32 version.

    image-classification31.7 MBLiteRT

    Develop on Arm with your AI coding assistant

    The Arm AI Portal MCP Server makes it faster to find optimized models, integrate code examples and deploy to your target device from your agentic coding assistant.

    User guide

    Run the following command in your terminal:

    codex mcp add arm-ai --url https://mcp.api.devplatform.arm.com/ai-portal