AI Briefing#multimodal-llm
The latest AI and developer news about #multimodal-llm, with the original source and a short summary.
The latest AI and developer news about #multimodal-llm, with the original source and a short summary.
The 1.6T parameter model achieves 88.4 on Terminal Bench v2.1 and 91.9 on BrowseComp, outperforming LongCat-2.0 and approaching frontier models like Claude-Opus-5.
Published in Nature, the 8B parameter EVLF-FM model demonstrates superior performance to human clinicians in medical imaging tasks.
A roundup of 10 latest AI/ML papers addressing agent efficiency and MCP governance issues.
The new model achieves a Decision Index score of 61.56, surpassing Jev by more than 3.5 points using the same Qwen3.8-27B backbone.
NVIDIA Dynamo has unveiled a serving architecture that separates vision encoders to reduce the TTFT of multimodal LLMs by up to 93%.
Z.ai released GLM-5.3-Flash with 18B active parameters, achieving performance close to Claude Opus 4.8 at one-tenth the cost.
Z.ai's GLM 5.3 Flash has launched on Vercel AI Gateway, enabling its use for coding and agentic tasks.
LangChain released a multi-vector retriever supporting diverse data types such as tables and images.
Apple proposed the IVT framework, which internalizes visual thinking during inference to reduce latency by more than 5x.
The study comparatively analyzed preference alignment methods and datasets for multimodal LLMs and proposed BDHS.
Moonshot AI has released the weights and technical documentation for Kimi K3, an MoE model with 2.78 trillion parameters.
Swiss-AI has released Apertus-v1.5, an open-source model equipped with multimodal capabilities and a thinking mode.
Thinking Machines has unveiled Inkling, a large-scale multimodal MoE model with 1 trillion parameters.
Horus Hiero, an open-source hieroglyph translation and multimodal model based on Qwen 3.5, has been released.
Gemma 4 is a next-generation open-weight multimodal model with diverse architectures and innovative features.
NVIDIA has released Audex-30B, a unified multimodal MoE-based model that processes text and audio simultaneously.
Qwen-RobotNav is a scalable unified model that can adjust observation context to fit diverse navigation tasks.
A fine-tuned model specialized for infographic structure output, based on SenseNova U1-8B-MoT, has been released.
China's MiniMax has released M3, a multimodal model that captures both performance and cost-efficiency through its innovative MSA architecture.
Qwen-VLA is a general-purpose VLA model that goes beyond visual understanding to perform continuous robot actions and trajectory prediction.
Apple will participate in CVPR 2026 to present its latest research, including generative AI and multimodal LLMs.
Ovis2.6-80B-A3B has been released as an MoE-based multimodal model.
It optimized multimodal data mixtures at 1/100 the cost using a small proxy model and Gaussian process.
The latest AI models excel at text-based financial analysis, but their performance drops sharply when analyzing image-based investor decks containing charts.