Is your app still calling deepseek-r1-distill-qwen-32b on cloudflare workers ai?
It stops answering on nov 20. so does nemotron-3-120b-a12b. fresh models are in the oven, time to switch 🍪
AMD CEO Lisa Su held up a $1,499 lunchbox-sized machine on stage and ran a 235-billion-parameter AI…
AMD CEO Lisa Su held up a $1,499 lunchbox-sized machine on stage and ran a 235-billion-parameter AI model on it live, with no data center or cloud behind it. It's a direct shot at Nvidia's $4,000 AI box. Inside is AMD's Ryzen AI Max+ 395, the first x86 chip whose CPU and GPU…
I'm hearing some rumblings about Deepseek changes that are already available...
I've been focused on Qwen 4 today but you better believe I've still been running Deepseek tests, and yeah I believe something big is going to be announced yet again. Just look at this!
Qwen4 max directly kills DeepSeek Flash (Harness).
For the same task and the same prompt, Qwen4's token is only 117k tokens, while DeepSeek's directly reaches 452k tokens, and the effect is not good! Thanks to @loktar00 for testing!
You have reached the end of the archive
All of deepseek