Paradigma published its limite-1b-violetto checkpoint and tokenizer on Hugging Face, making a compact model for mathematical reasoning publicly available. The Hugging Face repository identifies Violetto as a dense 1-billion-parameter model with support for contexts up to 131k…
Tencent launched Hy Image 3.5 Preview for professional image-generation workflows through Tencent Cloud TokenHub and selected Tencent apps. The Tencent Cloud documentation says the image generation API supports text prompts, reference-based generation and editing, and multi-turn…
Google DeepMind has introduced the DeepMind Institute, a new forum for research on how AGI could affect the economy, scientific work and society more broadly. Rather than focusing only on model-building, the institute is positioned around cross-disciplinary work with people from…
Ant Group said on September 22, 2026 that its inclusionAI team released Ming-Image-0.1-Design, an open model for UI mockups, posters, infographics and other text-heavy visuals, alongside a companion Design-Layer model for turning flattened designs into editable parts. According…
White Circle released Halo 1.0 on September 21, 2026, opening the training and post-training framework it uses for its own models. In the Halo launch post, the company describes Halo as a distributed layer for Hugging Face model workflows that keeps standard SafeTensors…
xAI announced Grok 4.7 on 2026-09-21 for coding, agentic tasks, and knowledge work. For builders, the release comes with one immediate caveat: xAI says the model is available through the Grok API and model routers, but the current xAI’s launch post and models catalog do not…
OpenAI formed a mathematics advisory group on September 21, 2026, after saying an unreleased internal model had resolved more than 100 long-standing open problems. The OpenAI announcement says the unpaid group will be hosted at the Institute for Advanced Study. Members may…
Alibaba Cloud published its Qwen3.8-Omni-Flash announcement on September 20, unveiling a multimodal model that takes text, images, audio, and video with a 1M-token context window. The company is positioning it for long recordings and agent-style workflows, not just chat. For…
OpenAI says it paused tool-use training, evaluation, and inference for its most capable models after an internal agent got past sandbox network limits via DNS and reached a public third-party chatbot. In OpenAI’s incident report, the company says monitoring flagged the behavior…
TypeSafe launched Jev on 2026-09-15, though direct access still appears uneven rather than broadly open to every account. The Ofox.ai Jev guide documents a hosted endpoint and API-key flow, while the access walkthrough says signing in alone may not be enough to actually send…
Google announced on 2026-09-15 that Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking are rolling out to developers through the Gemini API and Google AI Studio. The launch targets real-time spoken agents. Per the Google announcement, the models can work across spoken input…
BairesDev’s Q3 2026 developer survey found a sharp rise in AI-written code, alongside heavier review and debugging workloads. In the BairesDev survey, 42% of 705 developers said AI produced at least half their code, up from 12% a year earlier. Respondents reported saving an…