Model	techniques
GLM-5.3-Flash used	Reusing the image token for video frames used—
DeepSeek-V4.1-Flash core	3×3 pixel-unshuffle downsampling coreJoint vision-language pre-training coreMultimodal mixture-of-experts Transformer coreVision encoder–MLP projector pathway coreDisaggregated encoder training usedDiscarding the LLM after vision-encoder training usedHigh-resolution autoregressive vision-encoder fine-tuning usedLinear-projection patch embedding usedStaged vision-encoder freezing usedTwo-layer MLP vision projector used—
DeepSeek-V4-Flash-0731 core	Native multimodal visual understanding core—
MiMo-V2.6-Flash core	Four-frame audio patches with within-patch bidirectional self-attention coreJoint multimodal input and understanding coreTwo-stage audio encoding with tokenization and patch encoding coreData-parallel-first multimodal encoding usedJoint full-model multimodal training usedTwo-stage text-then-multimodal pre-training used—
MiMo-V2.5 core	Native multimodal understanding coreNative visual and audio understanding coreAudio encoder initialized from MiMo-Audio usedProjector warmup usedVision Transformer (ViT) usedVisual and audio encoders with lightweight projectors used—
MiniMax-M3 core	Mixed-modality training core—
DeepSeek-V4-Flash-Vision-Exp core	Native multimodal visual understanding coreVision encoder and multimodal aligner coreVisual modules for multimodal understanding core—
DeepSeek-V4-Pro-0813 core	Native multimodal visual understanding core—
Gemma 4 31B core	Encoder-free multimodal architecture coreDedicated multimodal encoders usedFrozen encoders during pre-training usedLightweight vision embedding module usedRaw audio projection into the LLM embedding space usedSingle-matrix-multiplication vision projection usedUniversal Speech Model (USM)-based audio encoder usedVariable aspect-ratio image handling usedConfigurable visual token budget optional—
Inkling core	Discrete token encoding for audio coreHierarchical patch encoder coreJoint multimodal decoding in a shared hidden space coreNative multimodal encoding coreEncoder-free multimodal architecture used—
Kimi K3 core	Single shared multimodal backbone core2×2 pixel-shuffle downsampling usedDecoupled Encoder Process (DEP) usedDual-format coordinate supervision usedFrom-scratch vision-encoder training with next-token prediction usedMoonViT-V2 usedMultimodal input used—
MiMo-V2.5-Pro core	Native multimodal understanding coreNative visual and audio understanding coreProjector warmup usedVisual and audio encoders with lightweight projectors used—
MiMo-V2.6-Pro core	Four-frame audio patches with within-patch bidirectional self-attention coreTwo-stage audio encoding with tokenization and patch encoding coreData-parallel-first multimodal encoding usedJoint full-model multimodal training usedTwo-stage text-then-multimodal pre-training used—
Qwen3.5-397B-A17B core	Early fusion multimodal training coreThinker–Talker architecture coreAudio Transformer (AuT) usedCausal streaming ConvNet codec decoder usedExplicit text-string timestamps usedMel-spectrogram audio front end usedResidual Vector Quantization (RVQ) speech representation usedVideo preprocessor longest-edge configuration optional—
Qwen3.6-35B-A3B core	Early fusion multimodal training core—
Step-3.7-Flash core	Language backbone coreVision encoder coreVision Transformer (ViT) used—

multimodal architecture

In this branch 57

filed here 57

By model