Liquid AI: vision encoders can burn more compute than the LLM on large images and video
JosephJacks_ · x · 2026-11-05
Liquid AI explains how its VLM team teaches LFMs to see: adding vision is more than one extra modality — with large images or long video, the vision encoder can use more compute than the language model itself, demanding a different model design. The team covers mid-training for vision and the custom encoder built for LFM3.
More from Models
- Swyx digs into MIT CSAIL's viral 'Weight-Space Programming': cool, reproducible, but temper the hype — rschu · 2026-11-05
- Together AI hosts live walkthrough of Kimi K3.1 Pro with Moonshot AI — togethercompute · 2026-11-05
- Model selection is a routing problem: match model size to task ambiguity, says Together — togethercompute · 2026-11-05
- Mistral Large 4 1T hits HF Inference Endpoints at $10/hr on 4x A100 — ngxson · 2026-11-05
- Small model 1017 loses evals to Haiku but "sometimes has frontier-level ideas," argues researcher — teortaxesTex · 2026-11-05
- Tencent's Hy4 benchmarks close in on Claude Opus 5.5 and GPT 6.1 Sol, topping several leaderboards — heypearlai · 2026-11-05