Deepseek V4 pro just broke the economics of frontier AI.
Claude Fable 5 is still smarter. Grok 4.6 is catching up fast. But one hidden number makes DeepSeek impossible to ignore for agents. The intelligence race: → Claude Fable 5 still wins the hardest reasoning and software
Dario Amodei says open source models are a distraction in AI.
When a new one drops, he doesn't ask whether it's open. He asks whether it's good. Q: "But what about open source? If you stopped investing in the models and open source caught up, people could swap in open source."
Gemini 3.7 Flash just cooked Grok 4.6, Opus 5, GPT-5.6, Fable 5, Kimi K3, and DeepSeek V4 Flash.
Here are some of the test runs using Gemini 3.7 Flash.
I have OpenWebUI as a chat interface for non-chat-agent friends.
It automatically provisions DS4 keys from LiteLLM. I even have MiaAI_lab ‘s 2 x DGX Spark Deepseek V4 Flash setup w/ Vision working inside of OpenWebUI’s chat interface (even though DS4 doesn’t natively have
You have reached the end of the archive
All of deepseek