model weights | Bit.Fan

Microsoft lists Mage-VL-Cowork model weights on Hugging Face

Techub News reported that the Microsoft organization has listed the Mage-VL-Cowork model weights on Hugging Face. Based on the information available, the weight upload may have appeared before any formal announcement from Microsoft. That timing has led to speculation that the company could be preparing an official release of the model. The report points to a newly posted weight entry on Hugging Face as the source of the development. No additional launch details, product documentation, or release timeline were disclosed in the brief update. The item remains limited to the appearance of the weights and the possibility that the listing preceded an official rollout.

NVIDIA

8 days agoNews Editor

NVIDIA lists MotionBricks model weights on Hugging Face

NVIDIA has listed a new set of model weights, "nvidia/MotionBricks," on Hugging Face, according to Techub. The appearance of model weights on the platform often comes before an official product announcement, and that appears to be the case here as well. As of now, the model has not been formally announced by NVIDIA. The update points only to the availability of the weights on Hugging Face and does not include any official launch statement from the company. Techub noted that early weight listings are not unusual and can precede a broader release or public introduction. At this stage, the listing itself is the main confirmed development.

550

China

14 days agoNews Editor

China plans export curbs on 23 key AI technologies, including model weights and source code

China is planning export controls on 23 key artificial intelligence technologies, according to a Techub News brief citing The Decoder. The proposed scope includes model weights, source code, and training methods for systems that reach 10^26 FLOPs in compute or 2 trillion parameters in scale. The report also says a fast-track rule would remain in place for open-source versions, allowing their release after registration is completed. That carve-out could limit some of the impact on open models, though the overall framework would still tighten oversight on advanced AI technology exports. The Decoder said the move may push Western companies to rely more heavily on U.S. open-source models instead of Chinese alternatives. No additional details on implementation timing were provided in the brief.

620

Meta to release model weights under Muse license with limited commercial-use terms

Meta Platforms (META.O) plans to release model weights under the Muse model license, according to ChainCatcher. The release will include limited commercial-use terms rather than fully unrestricted use. ChainCatcher said the framework will allow many developers to use the model free of charge. The brief update did not disclose additional technical details, timing beyond the report itself, or more specific licensing conditions. The announcement centers on access to the model weights and the scope of permitted use under the Muse license.

630

Qwen

22 days agoNews Editor

Qwen lists new Qwen-Image-2.1-Turbo weights on Hugging Face

Qwen has listed a new model weight, "Qwen/Qwen-Image-2.1-Turbo," on the model hosting platform Hugging Face, according to Techub News. The page information indicates that the model is related to image processing. The report also notes that model weight listings often appear before an official release. At this stage, the information cited comes from a new weight entry on Hugging Face, with no additional release details provided in the source. The update points to a possible upcoming formal launch, but the source stops at the listing itself and the model’s category as shown on the page.

560

NVIDIA

22 days agoNews Editor

NVIDIA appears to have listed weights for a new SSD object detection model on Hugging Face

Techub News reported that NVIDIA has uploaded a new model checkpoint under its Hugging Face organization, using the name "nvidia/agile_one_s_place_ssd_n17_24050." The listing is seen as a possible sign that NVIDIA is preparing to release a new SSD object detection model. The report did not say that the product had been officially announced. It only pointed to the newly posted model weights and noted that such uploads often appear before a formal release. The cited source for the information was Hugging Face. Based on the available details, the development remains an apparent pre-release signal rather than a confirmed launch announcement from NVIDIA.

530

Musubi

25 days agoNews Editor

Musubi launches PolicyLM-1.7B, a lightweight model for real-time content moderation

AI infrastructure company Musubi has introduced PolicyLM-1.7B, a lightweight decision model built for real-time content moderation, according to a Techub News update citing TechCrunch. The company also opened the model weights, making the release notable for developers and teams working on moderation systems. Musubi said the model is designed specifically for content review tasks, with the stated goal of improving both moderation efficiency and accuracy. No additional technical specifications, deployment details, or performance benchmarks were disclosed in the brief. The announcement centers on two confirmed points: the launch of PolicyLM-1.7B and the release of its weights. The model is positioned as a purpose-built tool for live moderation workflows rather than a general-purpose AI system.

370

NVIDIA

29 days agoNews Editor

NVIDIA lists PixelDiT2-ImageNet model weights on Hugging Face

NVIDIA has listed a new model checkpoint, "nvidia/PixelDiT2-ImageNet," on Hugging Face, according to Techub. The model page is already accessible on the AI community platform, even though the company has not issued a formal announcement yet. That timing stands out because checkpoint uploads often appear before an official product release. In this case, the listing points to a possible near-term launch of a new NVIDIA model. The update was attributed to Hugging Face page information cited by Techub, and no additional release details were disclosed in the source. At the time of publication, the key confirmed facts are limited to the live model-weight page and the absence of an official NVIDIA statement.

720

Total 11 items