China Dominates Top 10 Open-Source Models on HuggingFace. Not only is the West completely outclassed (Mistral, Llama), the Chinese models are also outstanding.
RESEARCH
-
Dividing 3D Assets via Generative Exploded Dynamics
By
–
BANG
— AK (@_akhaliq) 31 juillet 2025
Dividing 3D Assets via Generative Exploded Dynamics pic.twitter.com/cluSTzFXjTBANG Dividing 3D Assets via Generative Exploded Dynamics
-
KREA Flux Training Methodology and Technical Details
By
–
if you want to learn about how we trained KREA Flux, we prepared a detailed blog in the link below:
-
FLUX Krea Open-Weight Model Approaches FLUX Pro Quality
By
–
FLUX Krea ranks higher than previous open-weights FLUX models and gets close to FLUX Pro quality.
— KREA AI (@krea_ai) 31 juillet 2025
(results from internal evals, soon in @ArtificialAnlys) pic.twitter.com/Bwg1ySRic0FLUX Krea ranks higher than previous open-weights FLUX models and gets close to FLUX Pro quality. (results from internal evals, soon in @ArtificialAnlys
) -
Krea-1 Open-Source Image Model Released with SOTA Realism
By
–
today, we're releasing an open version of Krea-1.
— KREA AI (@krea_ai) 31 juillet 2025
FLUX Krea is a new SOTA open-source image model with incredible realism and aesthetics.
try it for free 👇 pic.twitter.com/OpqLFkygwTtoday, we're releasing an open version of Krea-1. FLUX Krea is a new SOTA open-source image model with incredible realism and aesthetics. try it for free
-
New AI Model Available on Hugging Face Platform
By
–
Try the model on our platform, deploy privately, or use it for research on @huggingface
. Find more details in our blog: -

Advanced Multimodal AI Model Excels Across Finance Healthcare Manufacturing Energy
By
–
It surpasses other models in its class at understanding and analyzing a wide range of visual and multilingual data across domains like finance, healthcare, manufacturing, and energy.
-

Using spec to fix GPT-4o sycophancy bug as trust anchor
By
–
7/ When GPT-4o showed excessive “sycophancy,” they used the spec to: • Identify it as a bug
• Roll back changes
• Clarify values
• Guide model updates The spec became a trust anchor. -

OpenAI’s Model Spec: A Plain English Markdown Document for Alignment
By
–
6/ OpenAI's solution? The Model Spec. A markdown doc:
• Written in plain English
• Versioned
• Open to policy, legal, safety teams
• Used to align both humans and AI It’s now central to how they build. -

Mistral Requires Comparison at a Different Scale
By
–
To compare Mistral's performance with other models, you need to change the scale entirely.
