Unverified50% confidenceFactExact time
The d1-omni-600M model has about 587 million parameters, is based on a small bidirectional encoder with added vision and audio encoders, supports text plus images or text plus up to 30 seconds of audio, and Liquid calls it an early research version.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/13/2026
First Seen
Valid until: 1/11/2027
Sources
Related Entities
Related Claims
Unverifiedd1-omni-600M 包含 381M 共享主干加决策头、94M SigLIP2 视觉编码器（来自 LFM2.5-VL-450M）和 112M 17 层 FastConformer 音频编码器。71% similarUnverifiedRunning a large language model with 70B parameters typically requires more than 40GB of VRAM (assuming 4-bit quantization), and video generation models also require more than 20GB of VRAM at high resolutions71% similarUnverifiedEmbeddingGemma 2's 740M parameters are divided into a text component of 270M (backbone 130M, embedding layer 140M), a vision encoder of 170M, and an audio encoder of 300M, with all three encoders loadable on demand.69% similarUnverifiedThe 26.2M model uses an 8-layer encoder-only Transformer, d_model 384, 6 attention heads, FFN dimension 1536, a maximum of 160 tokens, a self-developed 16k byte-level BPE tokenizer, and 15,000 edit labels.68% similarUnverifiedIn ComfyUI's default MiniMax H3 i2v workflow, fast audio generation can be achieved by disconnecting the load image node, setting the width and height to 32, and specifying the duration.67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1013593API
curl https://kongchang.com/api/v1/knowledge/claims/1013593MCP
get_claim(id=1013593)