Mage
Microsoft released Mage, a family of lightweight 4B-parameter multimodal models designed for training and deployment on consumer-grade hardware.
Deep dive
- The 4B-parameter budget is designed to keep training and fine-tuning viable on modest hardware.
- Mage-VL utilizes codec-inspired streaming to process video via anchor frames.
- Mage-Flow uses a VAE and native-resolution MMDiT for text-to-image tasks.
- The models emphasize 'codec-aligned' efficiency to focus compute on signal-heavy representations.
- Variants include base, RL-tuned, and Turbo versions for different latency requirements.
Decoder
- 4B-parameter budget: A constraint limiting the model to approximately 4 billion trainable weights, significantly smaller than state-of-the-art models like GPT-4 or Claude 3.5.
- Codec-native: An architecture that processes data (like video frames) using structures similar to video compression algorithms rather than standard frame sequences.
- MMDiT: Multimodal Diffusion Transformer, a neural network architecture capable of processing both image and text inputs.
Original article
Mage
A Lightweight, Research-Friendly Multimodal Model Family
Mage is a family of lightweight, research-friendly multimodal models built at a fixed 4B-parameter budget, sharing a codec-aligned efficiency philosophy — spend representation capacity where the signal is — across both visual understanding and generation. Both models are compact enough to train, fine-tune, and deploy on modest hardware, yet remain competitive with much larger open systems.
Vision–Language · Understanding
Mage-VL: An Efficient Codec-Native Streaming Multimodal Foundation Model
A codec-native, from-scratch VLM for image & video understanding — reads video the way a codec does (anchor/predicted frames, 16×16 patches), with bio-inspired proactive streaming.
Generation · Editing
Mage-Flow: An Efficient Native-Resolution Foundation Model for Image Generation and Editing
A compact 4B generative stack (Mage-VAE + Native-Resolution MMDiT) for text-to-image generation and instruction-based editing at native resolution, with Base / RL / 4-step Turbo variants.
Fresh dev news, AI-digested daily. Never miss a trend at devoured.fyi