Latest

This is the first 3D native city world model released by a major Chinese company, marking a significant step in AI's ability to model and understand urban environments.

量子位@量子位的朋友们Source

For users interested in audio-visual AI, this open-source release provides a new tool for generating and controlling video and sound, potentially at a lower cost than proprietary alternatives.

Primary sourcepandaily@[email protected] (Pandaily)Source

DeepSeek's new V4.1 Flash model is now available, replacing the V4 Pro version. The pricing for the new model started on September 10, and the transition to Flash is scheduled for September 14.

Primary sourcepandaily@[email protected] (Pandaily)Source

For users interested in 3D mapping and digital twins, ABot-Earth 0.7 represents a significant advancement in AI-generated 3D city models, offering a more efficient and immersive way to explore urban environments.

This matters because it shows how major AI companies are navigating complex regulatory hurdles, particularly regarding the security of their advanced models, which is a growing concern for the industry.

For developers working on humanoid robots, this open-source model provides a new foundation that can handle diverse tasks and generalize across different setups, potentially reducing the need for multiple specialized models.

For users interested in embodied AI, this is a significant update from a company that has been active in the space.

量子位@克雷西Source

Cost-effective for developers needing multimodal capabilities, with lower input activation costs than comparable models.

First open-source MoE model matching Qwen-Max performance, offering developers enterprise-grade capabilities without cost.

Primary sourcepandaily@[email protected] (Pandaily)Source

The 100x data scaling demonstrates significant progress in embodied AI with practical applications in robotics.

Primary sourcepandaily@[email protected] (Pandaily)Source

GPT-6 Astra's Looped Transformer architecture offers improved efficiency for long-context tasks compared to current models.

Primary sourceSebastian Raschka: Ahead of AI@Sebastian Raschka, PhDSource

The G1's autonomous sparring demonstrates significant progress in humanoid robotics without remote control.

Primary sourcepandaily@[email protected] (Pandaily)Source

Priced like V4 Flash but with native multimodal support, this limited beta offers developers a chance to test DeepSeek's new architecture before September 10.

Primary sourcepandaily@[email protected] (Pandaily)Source

MiniCPM5-2B offers superior performance for edge deployment with its 2B parameters, outperforming larger models while being accessible for developers in the region.

Available free on Ling Studio with BF16 and FP8 versions on Hugging Face and ModelScope, offering superior multimodal capabilities for developers in the region.

Developers can access this open-source model for autonomous driving applications, offering a cost-effective alternative to proprietary solutions.

Primary sourcepandaily@[email protected] (Pandaily)Source