Search results：Parameters

7 results

OpenAI's New Model Bel with 10 Trillion Parameters Completes Pre-training

OpenAI reportedly completed "Bel," a 10-trillion-parameter ultra-large pre-trained model, the next-generation foundation after "Doug," with major leaps in architecture and scale, targeting AGI. It excels in coding, reasoning, and long-horizon agent tasks.....

11 hr ago12.4K

Tencent Hunyuan Hy4 Preview Open Source: 770B Parameters, Million-Context, Standing at the Forefront of Open Source

Tencent releases Hunyuan Hy4 preview open-source model: 770B total/49B active params, 1M context. Official: pretraining & post-training co-advance, top open-source tier. Live on HuggingFace, GitHub, ModelScope, Gitcode; integrated with Tencent Cloud TokenHub & OpenRouter. Blind tests beat GLM-5.3 & Kimi K3 in productivity.....

3 days ago24.1K

Tencent Hunyuan launches open-source flagship model Hy4preview with 770B total parameters and 1M context

Tencent Hunyuan released the open-source flagship large model Hy4preview on August 28th, with a total of 770B parameters and 49B activated parameters, and a context length of 1M, positioning it as real productivity. The model has been open-sourced on HuggingFace, GitHub, ModelScope, and is also available on Tencent Cloud TokenHub and OpenRouter. It can be experienced through WorkBuddy/CodeBuddy, Yaobao, and IMA. It was built using expert data from fields such as software engineering, games, finance, and security.

3 days ago22.7K

Xiaopeng's Second-Generation VLA Large Model Receives First Major Upgrade, Comprehensive Safety Capabilities Increase by 20 Times

On Aug 27, XPeng unveiled a major upgrade to its second-gen VLA model at the 'TIME' AI event; XOS 6.3.0 debuts on G9L. The AI now understands time, advancing from static 3D to dynamic 4D spatiotemporal understanding, with larger on-device parameters to validate the Scaling Law for autonomous driving.....

3 days ago17.7K

Qwen3.8-Flash-Next Released, Early Spoiler of Qwen4 Architecture

Alibaba's Qwen team open-sourced Qwen3.8-Flash-Next, a multimodal MoE model previewing Qwen4. It has a highly sparse MoE design with 125B parameters, plus a 51B N-gram embedding table and 4B multi-token prediction module, delivering superior performance at very low compute cost.....

4 days ago15.0K

Handle Multiple Models in One Stop: OpenRouter Releases Asynchronous Video Generation API Integration Guide

Integrating video generation is easy, but switching models is painful due to inconsistent endpoints, parameters, statuses, polling, and outputs. OpenRouter's unified async video generation API standardizes workflows, simplifying development and reducing integration costs.....

5 days ago12.2K

Report: OpenAI's Secret Pre-training Bel Model with Over 10T Parameters Aims to Compete with GPT-6 and AGI Foundation

OpenAI has reportedly finished pretraining Bel, successor to Doug, with over 10T parameters, possibly the base model after GPT-6 and nearing AGI. AGI refers to human-like learning, reasoning, and adaptation across domains, unlike narrow AI.....

5 days ago11.7K