Union alpha lost to deepseek v4.1 flash just as I have expected.
Both models are multimodal, but Union alpha is 262k context window while DS v4.1 flash is 1m. the difference is clear. btw both models are available for FREE on cline desktop app rn.
DeepSeek-V4.1-Flash is beautifully crafted model that introduced so many bold takes on model architecture.
With CED, the model only activates 50% of the model during prefill which saves huge compute overhead, and CSA2 on top allows KV Cache reduction without sacrificing quality.
China's open source AI push isn't idealism.
It's aimed at the US labs' business model... DeepSeek v4 matched 98% of the top US models for 2% of the cost. Seven days after they shipped. No one pays $200 a month for that gap.
This was something else , something that took me long time.
A proof of concept that can be built on . what if you code your film not generate it ? made this this with cherry studio and deepseek 4.1 flash model .everything is made by procedurally . vectors , sounds folly. hope you
You have reached the end of the archive
All of deepseek