GPT-6 Astra is starting to show what the next generation of coding models looks like.

According to Arena, GPT-6 Astra (Max) is now #1 on Code Arena: WebDev, scoring:

• GPT-6 Astra (Max): 1797

• Claude Fable 5.1 (Max): 1762

• Claude Opus 5 (Max): 1688

That puts Astra 35 points ahead of the current #2, with particularly strong early results in areas like Data & Analytics, Consumer Products, and Content Creation Tools.

Another major Chinese AI model just joined ApiHub: Hunyuan Hy4 Preview.

Hy4 Preview is Tencent s new-generation MoE flagship model with:

• 770B total parameters

• 49B activated parameters per token

• 1M context length

If you tried Ox Alpha recently, here s the reveal: it was GLM-5.3-Flash.

GLM-5.3-Flash is the first natively multimodal model in the GLM-5 family, and it s now available on ApiHub.

Despite having 320B total parameters with only 18B activated, it delivers stronger overall capability than GLM-5.2 while being designed around much more efficient inference.

Built for efficiency

Ox Alpha has been getting a lot of attention recently, so we ve just added it to ApiHub.

And for now, you can use Ox Alpha for free.

No need to set up another model provider or maintain a separate integration you can access it through the same ApiHub API alongside DeepSeek, Qwen, GLM, MiniMax, and other models.

I m especially curious to see how developers use Ox Alpha in real projects:

When people talk about AI, the first thing that usually comes to mind is chatting with ChatGPT.

But AI can do much more than answer questions.

It can:

• Write and review code

• Extract information from documents

• Summarize reports and meetings

• Translate and rewrite content

• Analyze data

• Power customer support

• Automate repetitive workflows

• Call tools and APIs through agents

• Generate structured output for applications

• Become part of your own product

  

Two more models are now available on ApiHub:

• Qwen3.8-Max

• GLM-5.3

Both can be accessed through the same ApiHub integration, alongside models from DeepSeek, MiniMax, and other providers.

For developers working with multiple AI models, I m curious about something:

DeepSeek has updated V4-Pro to the new 0813 version.

DeepSeek just updated V4-Flash, and this one is especially interesting for developers.

The new DeepSeek-V4-Flash-0731 keeps the same architecture and model size as the preview version the improvements come from post-training.

The jump in agent performance is pretty significant:

It has been two weeks since I launched ApiHub on Product Hunt.

ApiHub provides one API for accessing multiple AI models, including affordable Chinese models such as DeepSeek, Qwen, GLM, and MiniMax...