Side effect of blocking Chinese firms from buying the best NVIDIA cards: top models are now explicitly being trained to work well on older/cheaper GPUs. The new SoTA model from @Kimi_Moonshot uses plain old BF16 ops (after dequant from INT4); no need for expensive FP4 support.
🚀 "Quantization is not a compromise — it's the next paradigm." After K2-Thinking's release, many developers have been curious about its native INT4 quantization format. 刘少伟, infra engineer at @Kimi_Moonshot and Zhihu contributor, shares an insider's view on why this choice…
Finally wrapped up the hardware half of Nand2Tetris, took me a week. It had been on my to-do list for way longer than I’d like to admit.
Deepseek engineers so cracked they bypassed cuda
United States 트렌드
- 1. Cyber Monday 39.4K posts
- 2. #Fivepillarstoken 1,549 posts
- 3. #IDontWantToOverreactBUT 1,189 posts
- 4. Alina Habba 19K posts
- 5. TOP CALL 10.8K posts
- 6. #MondayMotivation 9,115 posts
- 7. #GivingTuesday 2,318 posts
- 8. Shopify 3,919 posts
- 9. $MSTR 13.7K posts
- 10. Mainz Biomed N/A
- 11. Token Signal 3,085 posts
- 12. #Rashmer 17K posts
- 13. Check Analyze N/A
- 14. Market Focus 2,597 posts
- 15. Victory Monday 1,570 posts
- 16. Good Monday 42.2K posts
- 17. JUST ANNOUNCED 18.5K posts
- 18. Clarie 3,141 posts
- 19. World AIDS Day 17.5K posts
- 20. GreetEat Corp. N/A
Something went wrong.
Something went wrong.