Log inSign up
Image
user avatar
ModelScope
@ModelScope2022
Driving innovations with open communities. 💬 Join our Discord: discord.gg/hgKcfgXHAQ
HangZhou, China
modelscope.ai
Joined April 2024
157
Following
11.1K
Followers
RepliesArticlesMedia
  • user avatar
    ModelScope
    @ModelScope2022
    33m
    Turn Krea 2 into an inpainting and outpainting model. 🎨 INPAINTKREA is Aitrepreneur's specialized LoRA for editing or extending existing images. 🤖 modelscope.ai/models/Aitrepr… 🖌️ Built for masked-region regeneration and canvas expansion while preserving the surrounding image. 🧩
    Image
    Image
    Image
  • user avatar
    ModelScope
    @ModelScope2022
    20h
    1M-token context with only ~3B parameters active per token. ⚡ Meituan’s LongCat-Flash-Lite-Sparse brings sparse attention to long-context agents. 🤖 modelscope.ai/models/meituan… 🏆 The sparse variant scores 68.2 on SWE-Bench Verified and 59.33 on SWE-Bench Multilingual, up from 54.4
    Image
  • user avatar
    ModelScope
    @ModelScope2022
    23h
    Introducing Irodori-TTS-v4-Small from Aratako, a Japanese TTS model for multimodal voice design, zero-shot voice cloning, and Emoji-based Style Control. 🎙️ 🤖 modelscope.ai/models/Aratako… 🏆 With 120 seconds of reference audio, speaker-identification accuracy reaches 99.76% on JVS.
    Image
    00:00
  • user avatar
    ModelScope
    @ModelScope2022
    Aug 3
    🚨 Open-model community, make some noise! Qwen confirms that Qwen3.8-Max open weights are coming next week, alongside Qwen3.8-27B. 🚀 At 2.4T parameters, Qwen’s strongest model yet has demonstrated 10+ days of autonomous coding, thousand-step agent workflows, analysis of
    Image
  • user avatar
    ModelScope
    @ModelScope2022
    Aug 3
    2K. 15 seconds. Native stereo. One model. 🎬MiniMax-H3 is here, unifying video generation and editing across text, images, video, and audio. 🤖modelscope.ai/models/MiniMax… 🎥 Mix characters, camera motion, voices, styles, and other references in one natural-language prompt. ✏️
    Image
    00:00