yes. It's out. 169tok/s alone on A100x4 DeepSeek v4.1 Flash MTP 5
Built a 4-level demake of the 1989 Amiga classic Blood Money using DeepSeek V4.1 Flash!
👾 The entire project took less than an hour and cost just $2.03 total: 🎨 GPT-Image 2.5 (7 images): $0.27 🧠 DeepSeek V4.1 Flash: $1.76 Wild what you can build in an hour now. 🚀
Anthropic accuses Alibaba Qianwen, Dark Side of the Moon, Xiaomi, Zhipu, DeepSeek
Anthropic accuses Alibaba Qianwen, Dark Side of the Moon, Xiaomi, Zhipu, DeepSeek, SenseTime, and MiniMax of engaging in AI shelling, which is easier than manual nanotechnology. The tolerance of 0.00068 mm, hand-polished key components of domestic fighter aircraft, is not very reliable, because 680 nanometers is 0.68 microns, which can only theoretically achieve this accuracy, but it is more sincere than the Hanxin…
12GB RTX 4070 Ti graphics card vs 48GB Mac, who can carry the 27 billion parameter Qwen3.8 27B?
After watching several explanation videos, I still don’t understand the Quesma test report. I simply made an 11-minute instructional video: how to choose the six quantitative gears, how to install them, and what pitfalls there are, all in one go. Mentor GPT Astra, implement DeepSeek Flash 🎬…
You have reached the end of the archive
All of deepseek