• multimodal — A model that works with more than text — reading images, audio, or video, and sometimes generating them too.
• token — The chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.
• inference — Running a trained model to get answers — the phase where AI is actually used, as opposed to trained.

A 10B model reporting 92.2% on MMBench and 80.11% on MMMU narrows the gap to VLMs many times its size, making cheap or local multimodal inference viable, and the report documents the recipe behind it.

• postvLLM gains an experimental Attention-FFN disaggregation plugin for MoE
• postOpenRouter free tier for Step 3.5 Flash jumps to 10K calls a day
• postStep-Audio-R1.1 opens weights for a speech model that reasons in real time
• postNextStep-1.1 refines StepFun's image model with flow-based RL

• postSTEP3-VL-10B claims 200B-class multimodal results at 10B parametersStepFun
• postLFM2.5-VL-3B reads screens and grounds objects inside 3GBLiquid AI
• postMiMo-VL-7B posts GUI grounding numbers that beat models built for the jobXiaomi MiMo

Checking sign-in…

Loading comments…