Tencent Hunyuan releases Hy4 open weights
Tencent Hunyuan has released the open weights of Hy4, positioned as a native multimodal flagship MoE model with 770B…
The Multimodal topic tracks AI beyond text: image understanding and generation, audio and video input/output, cross-modal retrieval, and interaction. For readers following productization in vision, speech, and video, with links back to sources for verification.
Capabilities beyond text: models and product advances in image understanding, mixed text-image comprehension, and audio/video input and output.
Tencent Hunyuan has released the open weights of Hy4, positioned as a native multimodal flagship MoE model with 770B…
Arena shares a first-hand hands-on video following the release of Gemini 4 Neon, in which Peter Gostev compares it…
Google DeepMind opened a new Project Genie experience on November 18 to eligible Google AI Ultra subscribers, offering…
Tencent Hunyuan says Hy4 flash is now open-weight, including two versions: Hy4 flash-FP8 and Hy4 flash-Base. The model…
Google announced that Gemini 4 Hydrogen has rolled out in Google Search worldwide, calling it the largest model…
In the opening keynote of Ignite 2026, Satya Nadella announced multiple product updates for building a frontier…
Google releases Gemini 4 Carbon (gemini-4-carbon, 1M shared token pool, $2.00/$10.00 per 1M tokens until November 30.
Tencent Hunyuan launched Hy4 flash, with 224B total parameters (including 64B Engram parameters) and 12B activated,…
Qwen announced the release of Qwen4 Max. According to Artificial Analysis data, its Intelligence Index score is 43, 9…
World Labs announced that the Atlas world model is open to everyone, with Fei-Fei Li reposting and expressing…
Artificial Analysis evaluated Google's newly released Gemini 4 Hydrogen: the new xhigh reasoning tier scores 54 on the…
DeepSeek has opened in its API the capabilities previously used by its App chat service; developers can enable them…
One request can carry up to 1,000 questions. Google says results come back in about 150–250 milliseconds end to end —…
Google launched the Gemini Decisions API, powered by the new Gemini Decision 1 model, which can evaluate up to 1,000…
MiniMax has released H3-Image, an open-weight image model for text-to-image, reference-based generation and image…
After testing Claude Fable 5.2 against Gemini 4 Carbon, the author argues that Fable 5.2's standout difference lies in…
Baidu has released ERNIE 5.5 Flash Preview, the first post-trained version based on the earlier Base model, with 280B…
OpenRouter tested 15 vision models, Mistral OCR, and Tesseract on OCR performance using 8 self-made scanned pages with…
Anthropic officially released Claude Fable 5.2 in the early hours, and the author immediately tested 10 tasks…
An official Claude video introduces that Claude Mythos 5.2 predicted the complete AAV capsid structure, the protein…
Meituan's LongCat has open-sourced LongCat-2.5 under the MIT license, supporting commercial deployment and further…
Meta releases Muse Ember (muse-ember-1.0), positioned as its smallest and most efficient model for complex reasoning…
Anthropic released Claude Fable 5.2 and Claude Mythos 5.2, which are the same underlying model; Mythos requires access…
Ant Ling has released Ling-3.2-flash, a native multimodal model with 560B total parameters and 25B activated per step.…
Thinking Machines Lab has released Inkling 2 Small, its second-generation model. It is a 594B total-parameter MoE…
On Nov 5, 2026, Meta said real-time voice and expressive avatars are rolling out in the Muse app in the US on mobile,…
Google Research, together with Beth Israel Deaconess Medical Center and Harvard Medical School, published…
Preferred Networks released the translation-specialized LLM PLaMo 3 Translate, used in its own translation service, as…
Meituan's LongCat team has released LongCat-2.5, with 1.6 trillion total parameters and about 48 billion activated per…
Tencent Hunyuan released the official version of Hy4, expanding pretraining and post-training scale compared with Hy4…
Anthropic's biology research lab reported that with the help of Claude it obtained the first 3D structure of a vault…
Figma announced Figma Weave, a platform for creative teams with continuous output needs, connecting Weave creation…
Philipp Schmid relayed a message from Sundar Pichai: after launching a prototype satellite weeks ago, Google Project…
Moonshot (Kimi) released an official tutorial detailing the 3D creation workflow of Kimi K3.1 Pro, explaining…
OpenRouter released a tutorial on how to send a PDF as a file along with a strict JSON schema to a model that can read…
The vLLM team announced Day 0 support for Mistral's new flagship model Mistral Large 4, a 1T-parameter native…
ModelBest released and open-sourced the MiniCPM-OCR series. MiniCPM-OCR 1.0 (1.3B) achieved SOTA scores of 93.68 on…
Google releases the next-generation world model Genie 4, with clear improvements over Genie 3 in photorealism and…
Google Gemini announced the integration of SynthID detection: users can simply ask "Is this AI generated?" in Gemini…
Google DeepMind releases the world model Genie 4, featuring real-time interactive simulation, stronger photorealism…
Sponsored
Orange Relay (pro.nat6.net) is a premium AI API relay — an excellent transfer station with competitive pricing, reliable quality and broad model coverage including Claude, GPT and Gemini.
Visit Orange Relay