{ "schema": "https://ondevice.ai/schemas/recommendations.v1.json", "version": "2026-06-28", "status": "complete", "updatedAt": "2026-06-28T21:49:20Z", "updatedAtDisplay": "June 28, 2026, 22:49 BST", "site": { "name": "OnDevice.AI", "domain": "ondevice.ai", "api": "https://ondevice.ai/api", "description": "Current recommendations for open and open-weight AI models that can run comfortably on modern mobile devices with 8 GB of RAM." }, "targetDevice": { "ram": "8 GB", "class": "Latest high-end mobile phones", "ruleOfThumb": "Prefer 3B-4.5B effective dense language models in Q4_K_M, dedicated speech models below 1B parameters, and mobile-ready diffusion runtimes. Keep context modest on-device even when a model advertises very long context." }, "selectionNotes": [ "Filtered broad leaderboards, including Arena, for models that are realistically deployable on an 8 GB phone.", "Favored exact Hugging Face source repos and exact mobile runtime files over unofficial mirrors.", "Q4_K_M is the default recommendation for llama.cpp-compatible text and multimodal models because it is a strong quality-size tradeoff; speech and diffusion models use their native mobile formats instead.", "Kept the shortlist to permissive MIT or Apache-2.0 licenses where possible, including the image-generation pick." ], "recommendations": [ { "category": "best_text_generation", "categoryLabel": "Best text generation", "rank": 1, "model": { "displayName": "Qwen3.5 4B", "modelId": "Qwen/Qwen3.5-4B", "maker": "Qwen", "parameters": "4.66B", "license": "Apache-2.0" }, "why": "Best current Qwen-size balance for daily writing, chat, summarization, instruction following, multilingual text, and tool-style prompts.", "mobileFit": "Excellent. The Q4_K_M GGUF stays in the practical 8 GB phone envelope; use 8K-32K context on mobile rather than the full advertised long context.", "evidence": [ "Official model metadata reports 4.66B BF16 parameters and 262,144 native context length.", "Qwen reports strong current 4B-class scores, including MMLU-Pro 79.1, GPQA Diamond 76.2, IFEval 89.8, and LongBench v2 50.0." ], "original": { "format": "Transformers safetensors", "repo": "https://huggingface.co/Qwen/Qwen3.5-4B", "download": "https://huggingface.co/Qwen/Qwen3.5-4B/tree/main" }, "runtime": { "format": "GGUF", "recommendedQuantization": "Q4_K_M", "repo": "https://huggingface.co/unsloth/Qwen3.5-4B-GGUF", "download": "https://huggingface.co/unsloth/Qwen3.5-4B-GGUF/resolve/main/Qwen3.5-4B-Q4_K_M.gguf", "notes": "Use llama.cpp, LM Studio, Ollama-compatible import flows, or a mobile llama.cpp binding; disable thinking for quick instruction-style replies." } }, { "category": "best_coding", "categoryLabel": "Best coding", "rank": 1, "model": { "displayName": "Qwen3.5 4B", "modelId": "Qwen/Qwen3.5-4B", "maker": "Qwen", "parameters": "4.66B", "license": "Apache-2.0" }, "why": "The strongest small open model in this shortlist for coding tasks, with a newer unified Qwen3.5 base that can run with or without thinking.", "mobileFit": "Good. It fits the same Q4_K_M mobile envelope, but long thinking traces can consume time and context; cap output tokens for mobile apps.", "evidence": [ "Qwen reports LiveCodeBench v6 55.8 and OJBench 24.1 for Qwen3.5 4B.", "It also keeps strong instruction and function-style behavior in a compact Apache-2.0 model." ], "original": { "format": "Transformers safetensors", "repo": "https://huggingface.co/Qwen/Qwen3.5-4B", "download": "https://huggingface.co/Qwen/Qwen3.5-4B/tree/main" }, "runtime": { "format": "GGUF", "recommendedQuantization": "Q4_K_M", "repo": "https://huggingface.co/unsloth/Qwen3.5-4B-GGUF", "download": "https://huggingface.co/unsloth/Qwen3.5-4B-GGUF/resolve/main/Qwen3.5-4B-Q4_K_M.gguf", "notes": "Keep context and max output bounded on phones; enable thinking only when latency is acceptable." } }, { "category": "best_reasoning", "categoryLabel": "Best reasoning", "rank": 1, "model": { "displayName": "Qwen3.5 4B", "modelId": "Qwen/Qwen3.5-4B", "maker": "Qwen", "parameters": "4.66B", "license": "Apache-2.0" }, "why": "A compact current-generation reasoning pick with unusually high small-model math, logic, and coding benchmark scores.", "mobileFit": "Good for deliberate answers. It is not the fastest mode, but it is the best reasoning pick that still fits the 8 GB phone target.", "evidence": [ "Qwen reports GPQA Diamond 76.2, HMMT Feb 2025 74.0, and HMMT Nov 2025 76.8 for Qwen3.5 4B.", "The model supports thinking by default and can be configured for non-thinking responses when latency matters." ], "original": { "format": "Transformers safetensors", "repo": "https://huggingface.co/Qwen/Qwen3.5-4B", "download": "https://huggingface.co/Qwen/Qwen3.5-4B/tree/main" }, "runtime": { "format": "GGUF", "recommendedQuantization": "Q4_K_M", "repo": "https://huggingface.co/unsloth/Qwen3.5-4B-GGUF", "download": "https://huggingface.co/unsloth/Qwen3.5-4B-GGUF/resolve/main/Qwen3.5-4B-Q4_K_M.gguf", "notes": "Use thinking mode for deliberate answers; expose a fast non-thinking mode for quick UX." } }, { "category": "best_image_generation", "categoryLabel": "Best image generation", "rank": 1, "model": { "displayName": "LCM Dreamshaper v7", "modelId": "SimianLuo/LCM_Dreamshaper_v7", "maker": "Simian Luo", "parameters": "SD 1.5-class latent consistency pipeline", "license": "MIT" }, "why": "Best permissively licensed mobile-class image generation pick: fast few-step generation, MIT license, and a smaller SD 1.5-class footprint.", "mobileFit": "Good with ONNX, Core ML conversion, or platform-specific diffusion runtimes. GGUF is not the right format for diffusion models.", "evidence": [ "Hugging Face lists LCM Dreamshaper v7 as an MIT-licensed Diffusers text-to-image model.", "The repo includes ONNX assets for the text encoder, UNet, VAE encoder, and VAE decoder, making it a practical mobile runtime target.", "It trails SD Turbo on quality and prompt following, but avoids the non-commercial SAI license." ], "original": { "format": "Diffusers safetensors", "repo": "https://huggingface.co/SimianLuo/LCM_Dreamshaper_v7", "download": "https://huggingface.co/SimianLuo/LCM_Dreamshaper_v7/resolve/main/LCM_Dreamshaper_v7_4k.safetensors" }, "runtime": { "format": "ONNX / Core ML / Diffusers", "recommendedQuantization": "N/A", "repo": "https://huggingface.co/SimianLuo/LCM_Dreamshaper_v7", "download": "https://huggingface.co/SimianLuo/LCM_Dreamshaper_v7/tree/main", "companionFiles": [ "https://huggingface.co/SimianLuo/LCM_Dreamshaper_v7/resolve/main/unet/model.onnx", "https://huggingface.co/SimianLuo/LCM_Dreamshaper_v7/resolve/main/vae_decoder/model.onnx", "https://huggingface.co/SimianLuo/LCM_Dreamshaper_v7/resolve/main/text_encoder/model.onnx" ], "notes": "Use 4-step LCM-style generation for mobile UX; do not look for Q4_K_M GGUF for this category." } }, { "category": "best_speech_to_text", "categoryLabel": "Best speech-to-text", "rank": 1, "model": { "displayName": "Whisper Large V3 Turbo", "modelId": "openai/whisper-large-v3-turbo", "maker": "OpenAI", "parameters": "809M", "license": "MIT" }, "why": "Best dedicated open ASR choice for quality, language coverage, and mobile deployability after quantization.", "mobileFit": "Good. Use whisper.cpp q5_0 or a Core ML encoder package for phone deployment.", "evidence": [ "Hugging Face lists the model under automatic-speech-recognition with MIT license and around 809M parameters.", "The whisper.cpp repo provides ggml-large-v3-turbo q5_0 and Core ML encoder artifacts." ], "original": { "format": "Transformers safetensors", "repo": "https://huggingface.co/openai/whisper-large-v3-turbo", "download": "https://huggingface.co/openai/whisper-large-v3-turbo/resolve/main/model.safetensors" }, "runtime": { "format": "whisper.cpp GGML", "recommendedQuantization": "q5_0", "repo": "https://huggingface.co/ggerganov/whisper.cpp", "download": "https://huggingface.co/ggerganov/whisper.cpp/resolve/main/ggml-large-v3-turbo-q5_0.bin", "companionFiles": [ "https://huggingface.co/ggerganov/whisper.cpp/resolve/main/ggml-large-v3-turbo-encoder.mlmodelc.zip" ], "notes": "Whisper uses GGML/whisper.cpp formats rather than GGUF." } }, { "category": "best_text_to_speech", "categoryLabel": "Best text-to-speech", "rank": 1, "model": { "displayName": "Kokoro 82M", "modelId": "hexgrad/Kokoro-82M", "maker": "hexgrad", "parameters": "82M", "license": "Apache-2.0" }, "why": "Very small, popular, high-quality open TTS with broad community runtime support and many voice files.", "mobileFit": "Excellent. The ONNX q4/q8 variants are tiny compared with LLMs and are a better mobile target than GGUF.", "evidence": [ "Hugging Face lists Kokoro 82M as Apache-2.0 text-to-speech with very high downloads and likes.", "The ONNX community conversion includes model_q4.onnx, model_q8f16.onnx, and voice binaries." ], "original": { "format": "PyTorch checkpoint", "repo": "https://huggingface.co/hexgrad/Kokoro-82M", "download": "https://huggingface.co/hexgrad/Kokoro-82M/resolve/main/kokoro-v1_0.pth" }, "runtime": { "format": "ONNX", "recommendedQuantization": "model_q4.onnx or model_q8f16.onnx", "repo": "https://huggingface.co/onnx-community/Kokoro-82M-v1.0-ONNX", "download": "https://huggingface.co/onnx-community/Kokoro-82M-v1.0-ONNX/resolve/main/onnx/model_q4.onnx", "companionFiles": [ "https://huggingface.co/onnx-community/Kokoro-82M-v1.0-ONNX/resolve/main/voices/af_heart.bin" ], "notes": "Pick a voice binary alongside the ONNX model." } }, { "category": "best_photo_analyzer", "categoryLabel": "Best photo analyzer", "rank": 1, "model": { "displayName": "Gemma 4 E4B Instruct", "modelId": "google/gemma-4-E4B-it", "maker": "Google DeepMind", "parameters": "4.5B effective / 8B total with embeddings", "license": "Apache-2.0" }, "why": "Best verified small multimodal model for image understanding, OCR-like tasks, charts, screenshots, and general photo analysis.", "mobileFit": "Good on high-end phones with Q4_K_M. Vision requires the GGUF model plus an mmproj file, and lower visual token budgets improve speed.", "evidence": [ "Gemma 4 E4B is documented as targeting high-end phones, with text, image, and audio inputs and 128K context.", "Google reports MMMU Pro 52.6, MATH-Vision 59.5, OmniDocBench 0.181, and MMMLU 76.6 for the E4B instruction-tuned model." ], "original": { "format": "Transformers safetensors", "repo": "https://huggingface.co/google/gemma-4-E4B-it", "download": "https://huggingface.co/google/gemma-4-E4B-it/tree/main" }, "runtime": { "format": "GGUF + mmproj", "recommendedQuantization": "Q4_K_M", "repo": "https://huggingface.co/bartowski/google_gemma-4-E4B-it-GGUF", "download": "https://huggingface.co/bartowski/google_gemma-4-E4B-it-GGUF/resolve/main/google_gemma-4-E4B-it-Q4_K_M.gguf", "companionFiles": [ "https://huggingface.co/bartowski/google_gemma-4-E4B-it-GGUF/resolve/main/mmproj-google_gemma-4-E4B-it-f16.gguf" ], "notes": "Use the mmproj file for image input in llama.cpp-style runtimes." } }, { "category": "best_overall", "categoryLabel": "Best overall", "rank": 1, "model": { "displayName": "Gemma 4 E4B Instruct", "modelId": "google/gemma-4-E4B-it", "maker": "Google DeepMind", "parameters": "4.5B effective / 8B total with embeddings", "license": "Apache-2.0" }, "why": "Best all-around on-device choice because it covers text, reasoning, coding, images, and short audio input in one mobile-targeted model.", "mobileFit": "Good for flagship 8 GB phones when quantized; choose Qwen3.5 4B if your app is text-only and needs stronger writing, coding, or reasoning.", "evidence": [ "Gemma 4 E4B is explicitly positioned for mobile and edge deployment.", "Google reports E4B benchmark strength across text, coding, reasoning, vision, audio, and multilingual tasks." ], "original": { "format": "Transformers safetensors", "repo": "https://huggingface.co/google/gemma-4-E4B-it", "download": "https://huggingface.co/google/gemma-4-E4B-it/tree/main" }, "runtime": { "format": "GGUF + mmproj", "recommendedQuantization": "Q4_K_M", "repo": "https://huggingface.co/bartowski/google_gemma-4-E4B-it-GGUF", "download": "https://huggingface.co/bartowski/google_gemma-4-E4B-it-GGUF/resolve/main/google_gemma-4-E4B-it-Q4_K_M.gguf", "companionFiles": [ "https://huggingface.co/bartowski/google_gemma-4-E4B-it-GGUF/resolve/main/mmproj-google_gemma-4-E4B-it-f16.gguf" ], "notes": "This is the general default when one model must cover multiple on-device experiences." } } ], "sources": [ { "label": "Arena leaderboard", "url": "https://arena.ai/leaderboard/" }, { "label": "Hugging Face Models", "url": "https://huggingface.co/models" } ], "history": [ { "version": "2026-06-28", "updatedAt": "2026-06-28T21:49:20Z", "html": "/archive/2026-06-28/", "api": "/archive/2026-06-28/api.json", "notes": "First complete OnDevice.AI shortlist for 8 GB mobile devices." } ] }