Apple

• LiTo Gen 3 Multimodal

  Released 6mo ago

  Apple’s research image-to-3D model and tokenizer for generating 3D objects with unified geometry and view-dependent appearance.

  • 🧊 3D
  • 🌍 3D images
  • 🎲 3D objects
• Manzano Gen 3 Multimodal

  Released 7mo ago

  Apple’s unified multimodal model that shares a hybrid vision tokenizer for both image understanding and text-to-image generation, using one autoregressive LLM plus a diffusion decoder to reach state-of-the-art unified p…

  • 🖼️ Image generation
  • 🔊 Text to speech
  • 🔍 SEO content
  • 🎮 Game creation
• SHARP Gen 4 Image

  Released 9mo ago

  Apple’s monocular view-synthesis model that regresses a 3D Gaussian scene from one photo in under a second on a standard GPU, enabling real-time, photorealistic nearby views with metric camera motion.

  • 🖼️ Image generation
  • 🔍 Image recognition
• FastVLM Gen 3 Text

  Released 1y ago

  Apple’s lightweight vision-language model built for real-time multimodal apps. It ingests images alongside text and returns grounded answers fast—OCR, charts/diagrams, screenshots, and general visual QA—while supporting…

  • 📜 OCR
  • 🔍 SEO content
  • 📞 Customer support
  • 📊 Data analysis
• Image Playground Gen 4 Image

  Released 1y ago

  Apple’s on-device generative image tool inside Apple Intelligence. It creates playful visuals in three styles—Animation, Illustration, and Sketch—right inside apps like Messages and Notes, and is also available to devel…

  • 📷 Images
  • 🔊 Text to speech
  • 🖼️ Image generation
  • 🍸 Cocktail recipes
• MM1.5 Gen 3 Text

  Released 1y ago

  Apple Research’s refinement of the MM1 multimodal recipe. It keeps the same encoder–decoder architecture but upgrades data curation, image resolution, and multi-image/document training, yielding stronger OCR, layout und…

• MM1 Gen 3 Text

  Released 2y ago

  Apple Research’s multimodal LLM blueprint: a vision encoder feeding a text decoder via cross-attention, pretrained on a balanced mix of image–caption, interleaved image–text, and text-only data. It highlights how data q…

Fundraises

• Apple - Debt Financing Announced 30 Apr 2013 · $17B round · 0 investors · Closed
• Apple - IPO Announced 12 Dec 1980 · $101.20M round · 0 investors · Closed
• Apple - Seed Announced 15 Jan 1977 · $250K raised · 1 investors · Closed
• Apple - Series A Announced 15 Jan 1978 · $517.50K raised · 3 investors · Closed

News

• Apple acquires brain imaging firm for health and accessibility features Apple has bought Sonera, a California company that makes noninvasive sensors that can measure neural signals…

Papers

• Application-Agnostic Language Modeling for On-Device ASR On-device automatic speech recognition systems face several challenges compared to server-based systems. They…
• Differentially Private Heavy Hitter Detection using Federated Analytics In this work, we study practical heuristics to improve the performance of prefix-tree based algorithms for di…
• Diffusion Models Without Attention In recent advancements in high-fidelity image generation, Denoising Diffusion Probabilistic Models (DDPMs) ha…
• Efficient Large Language Model Inference with Limited Memory Tackles running LLMs that don’t fit DRAM by keeping weights in flash and streaming them efficiently. Introduc…
• Evaluating Discourse Phenomena in Neural Machine Translation For machine translation to tackle discourse phenomena, models must have access to extra-sentential linguistic…
• Federated Evaluation and Tuning for On-Device Personalization: System Design & Applications We describe the design of our federated task processing system. Originally, the system was created to support…
• High Speed All-optical extended DV-Curve-based DNA sequence alignment utilizing wavelength and polarization modulation This paper presents a novel optical processing approach for exploring genome sequences built upon optical cor…
• Learning from Simulated and Unsupervised Images through Adversarial Training With recent progress in graphics, it has become more tractable to train models on synthetic images, potential…
• Learning with Privacy at Scale Foundational Apple paper describing a system architecture for learning from user populations with local diffe…

Repositories

• ARKitScenes This repo accompanies the research paper, ARKitScenes - A Diverse Real-World Dataset for 3D Indoor Scene Unde…
• axlearn An Extensible Deep Learning Library
• coreai-optimization A library for PyTorch model compression and optimizations for deployment via Core AI on Apple silicon.
• coremltools Core ML tools contain supporting tools for Core ML model conversion, editing, and validation.
• ml-4m 4M: Massively Multimodal Masked Modeling
• ml-ane-transformers Reference implementation of the Transformer architecture optimized for Apple Neural Engine (ANE)
• ml-batchquant BatchQuant: Quantized-for-all Architecture Search with Robust Quantizer
• ml-direct2.5 Direct2.5: Diverse Text-to-3D Generation via Multi-view 2.5D Diffusion
• ml-encqa Repository for the research paper: EncQA: Benchmarking Vision-Language Models on Visual Encodings for Charts…
• ml-famae bioFAME: Frequency-Aware Masked Autoencoders for Multimodal Pretraining on Biosignals
• ml-health-query-profiles Query Profile
• ml-hypersim Hypersim: A Photorealistic Synthetic Dataset for Holistic Indoor Scene Understanding
• ml-icr2 ICR2: Eliciting In-context Retrieval and Reasoning for Long-context Large Language Models
• ml-isqoe Official implementation of the paper
• ml-l3m Large multi-modal models (L3M) pre-training.
• ml-lvsum-video-summarization LVSum: A Benchmark for Timestamp-Aware Long Video Summarization
• ml-matrix3d [CVPR 2025 Highlight] Matrix3D: Large Photogrammetry Model All-in-One
• ml-mkqa We introduce MKQA, an open-domain question answering evaluation set comprising 10k question-answer pairs alig…
• ml-nvas3d Novel-View Acoustic Synthesis from 3D Reconstructed Rooms
• ml-perceptual-3dgs Drop-In Perceptual Optimization of 3D Gaussian Splatting
• ml-pfl4asr Private Federated Learning for Speech Recognition
• ml-policy-projector Policy Maps (UIST 2025)
• ml-qrecc Open-Domain Question Answering Goes Conversational via Question Rewriting
• ml-quant Research publication code for
• ml-rl-universal-sequence-design ml-rl-universal-sequence-design

Datasets

• DataCompDR-1B Dataset · License apple-amlr · text to image · 41,214 downloads · by apple

Demos

• fastvlm-webgpu Real-time video captioning powered by FastVLM · running · License unknown

Devices

• Mac mini (M6 / M5 Pro) Ultracompact desktop computer optimized for on-device AI, local LLMs, AI agents, creative workloads and profe…

• Debt Financing Announced 30 Apr 2013
  $17B Round Closed
• IPO Announced 12 Dec 1980
  $101.20M Round $1.78B Post-money Closed
• Series A Announced 15 Jan 1978
  $517.50K Raised $517.50K Round $5.20M Post-money 3 investors Closed
• Seed Announced 15 Jan 1977
  $250K Raised $250K Round $750K Pre-money 1 investors Closed