Elon fait le bon choix en open-sourçant l’IA. En rendant publiques les anciens modèles de Grok, il accélère aussi l’innovation en IA. Anthropic reste aujourd’hui le laboratoire d’IA le plus fermé.
OPEN SOURCE
-
Microsoft VibeVoice-1.5B: Open-Source Multilingual Speech Generation
By
–
🎙️ Microsoft just dropped VibeVoice-1.5B — an open-source speech model that can:
— Futurepedia – Learn to Leverage AI (@futurepedia_io) 28 août 2025
▫️Generate 90 minutes of audio
▫️Up to 4 unique voices
▫️Expressive
▫️Multilingual
▫️MIT-licensed.
A game-changer for voice AI.
👉 For more AI breakthroughs, follow @futurepedia_io
Swing by… pic.twitter.com/5lZ1weu6feMicrosoft just dropped VibeVoice-1.5B — an open-source speech model that can: Generate 90 minutes of audio Up to 4 unique voices
Expressive
Multilingual
MIT-licensed. A game-changer for voice AI. For more AI breakthroughs, follow @futurepedia_io Swing by -
Prime Intellect Launches Collaborative Environments Hub for RL AGI
By
–
Prime Intellect launched Environments Hub, an open platform for crowdsourcing environments where agents learn
— The Rundown AI (@TheRundownAI) 28 août 2025
It allows users to create, explore, and reuse environments, lowering the barrier for anyone to contribute to state-of-the-art RL and AGIpic.twitter.com/PsmFi8oQ1xPrime Intellect launched Environments Hub, an open platform for crowdsourcing environments where agents learn It allows users to create, explore, and reuse environments, lowering the barrier for anyone to contribute to state-of-the-art RL and AGI
-
Google DeepMind Perch v2 Identifies Threatened Species from Audio
By
–
And it continues at Google! Definitely… Google DeepMind has just launched Perch v2, an open-source AI model capable of analyzing nature sounds to identify threatened species from audio recordings. Here's how it's already been used: – Discovery of a new population of the rare
-
Nous Research Introduces Hermes 4 Open Hybrid Reasoning Models
By
–
Nous Research introduced Hermes 4, a family of open, hybrid reasoning models Designed to be neutral, the models achieve SOTA on RefusalBench, which tests for willingness to be helpful in scenarios usually disallowed by both closed and open models
-
Tencent Launches HunyuanVideo-Foley Open-Source Audio Generation AI
By
–
Tencent launched HunyuanVideo-Foley, an open-source Text-Video-to-Audio AI
— The Rundown AI (@TheRundownAI) 28 août 2025
Trained on 100k-hours of data, it generates contextually-aware soundscapes for scenes ranging from natural landscapes to animated shorts
Also achieves SOTA on multiple benchmarkspic.twitter.com/0CWII2KWyeTencent launched HunyuanVideo-Foley, an open-source Text-Video-to-Audio AI Trained on 100k-hours of data, it generates contextually-aware soundscapes for scenes ranging from natural landscapes to animated shorts Also achieves SOTA on multiple benchmarks
-
MiniCPM-V 4.5 Chat App Development with Impressive Benchmark Performance
By
–
vibe coding a MiniCPM-V 4.5 @OpenBMB chat app in anycoder
— AK (@_akhaliq) 28 août 2025
MiniCPM-V 4.5 achieves an average score of 77.0 on OpenCompass, a comprehensive evaluation of 8 popular benchmarks. With only 8B parameters, it surpasses widely used proprietary models like GPT-4o-latest, Gemini-2.0 Pro,… pic.twitter.com/r1i5b6JfFpvibe coding a MiniCPM-V 4.5 @OpenBMB chat app in anycoder MiniCPM-V 4.5 achieves an average score of 77.0 on OpenCompass, a comprehensive evaluation of 8 popular benchmarks. With only 8B parameters, it surpasses widely used proprietary models like GPT-4o-latest, Gemini-2.0 Pro,
-
RAG-Anything: Open Source Multimodal RAG Framework
By
–
All-in-One RAG System! RAG-Anything is a unified framework with a multi-stage multimodal pipeline that extends traditional RAG architectures. It handles diverse content through intelligent orchestration and cross-modal understanding. 100% Open Source
-
AI Kosh Platform Launches Multilingual Models for India
By
–
The AI Kosh platform now hosts multilingual AI models developed by AI4Bharat, designed to empower applications with India’s linguistic richness: MultiIndic Question Generation A sequence-to-sequence multilingual model trained on a large corpus of 770K examples, also
-
Alibaba Releases Wan2.2-S2V Open-Source Speech-to-Video Model
By
–
Alibaba just dropped Wan2.2-S2V, a 14B parameter speech-to-video model
— The Rundown AI (@TheRundownAI) 27 août 2025
The AI delivers film-grade, audio-driven videos with dynamic consistency, advanced motion, and environment control
And, it's fully open-sourcepic.twitter.com/S5QgPrb43hAlibaba just dropped Wan2.2-S2V, a 14B parameter speech-to-video model The AI delivers film-grade, audio-driven videos with dynamic consistency, advanced motion, and environment control And, it's fully open-source