ModelBest open-sources MiniCPM-V 4.7, a 35B model in 3GB on a phone
ModelBest and OpenBMB release MiniCPM-V 4.7, including a 35B-A3B MoE that reportedly runs in about 3GB of memory on an iPhone 17 Pro Max at over 13 tokens/s.
multimodal AI News and Analysis
15 verified stories covering multimodal AI, product updates and industry developments.
Cloudflare's Clef-omni Decision Model Now Handles Video and Audio
Cloudflare's new Clef-omni scores audio and full video input in a single pass: about 1.5 seconds for a 21-second clip, at $0.15 per million input tokens.
Alibaba's New Live-Translation Model Cuts Delay to 2.3 Seconds
Qwen3.8-LiveTranslate trims per-character latency from 2.8 to 2.3 seconds, understands 60 languages and speaks 29, with a demo page and API now live.
World Labs launches Atlas, a controllable 1440p video model
World Labs' new Atlas model bakes camera pose into its conditioning inputs, unifying 3D reconstruction, camera-controlled video generation, and spatial simulation in one architecture.
WeChat's Open-Source Multimodal Embedding Model Tops MMEB-v2
Tencent's WeChat team open-sourced WeMM-Embedding, a multimodal embedding model whose 2B version tops the MMEB-v2 leaderboard, beating rivals four times its size.
DeepSeek launches experimental vision model, caps images at 384 tokens
DeepSeek's new deepseek-v4-flash-vision-exp caps every image at 384 tokens with no extra vision fee, effectively putting a ceiling on GUI agent screenshot costs.
SenseTime Open-Sources U1.5, 8B Model Outputs Native 4K
SenseTime open-sourced SenseNova-U1.5-8B-MoT with native 4K generation and precise editing, but its parameter count is reported three different ways.
Inkling Small Closes In on 975B Flagship
Thinking Machines released Inkling-Small with open weights, 276B total parameters, 12B active parameters, multimodal input and benchmark scores close to the 975B Inkling.
MiniMax Opens Its H3 Video Model
MiniMax H3 combines 2K, 15-second video, native stereo audio, RMB 0.8-per-second pricing and a coming open-weight release.
ByteDance targets design workflows with Seedream 5.0 Pro
ByteDance targets design workflows with Seedream 5.0 Pro is reframed for global technology readers, with the key numbers, product claims and open questions kept intact.
Alibaba’s Qwen3.7-Plus adds vision and agent loops
The multimodal version of Qwen3.7-Max can read images and video, but Alibaba is emphasizing autonomous iteration, tool use and self-testing.
Alibaba pitches Qwen3.7-Plus as a self-checking agent model
Alibaba released Qwen3.7-Plus on June 2, highlighting multimodal input, tool use, code generation and autonomous testing as its next agent bet.
Alibaba Launches Qwen3.5-Omni: Unified Text, Audio, Video Model with Voice Cloning
Alibaba's Qwen team has released Qwen3.5-Omni, its first true all-modal model that processes text, images, audio, and video within a single architecture, featuring real-time voice cloning and audio-video vibe coding.
Qwen3.5 supports 201 languages as Alibaba accelerates global push
Alibaba's Qwen3.5 open-weight model now covers 201 languages and dialects, introduces native multimodal processing, and signals a broader industry shift from chatbots to AI agents.
Kimi K2.5 Breaks Down Complex Tasks With 100 Agents
Moonshot AI's Kimi K2.5 introduces native multimodal understanding and an Agent Swarm feature that splits complex tasks into parallel sub-tasks handled by multiple agents, significantly boosting speed.