Hackernews posts about Flash Attention 4

1. 01
   

   newShow HN: LoRA over GGUF – Train Qwen3.8-Flash-Next in 40G VRAMgithub.com

   ▲ 1woctordho27ddiscusssavesaved
2. 02
   

   newWe reverse-engineered Flash Attention 4modal.com

   ▲ 134birdculture405d💬 48savesaved
3. 03
   

   newFlashAttention-4: Algorithm and Kernel Pipelining Co-Designresearch.colfax-intl.com

   ▲ 6skidrow125ddiscusssavesaved
4. 04
   

   newWe reverse-engineered Flash Attention 4modal.com

   ▲ 5charles_irl406ddiscusssavesaved
5. 05
   

   newFlashAttention-4: Algorithm and Kernel Pipeliningresearch.colfax-intl.com

   ▲ 4skidrow126ddiscusssavesaved
6. 06
   

   newMaking FlashAttention-4 faster for inferencemodal.com

   ▲ 4birdculture146ddiscusssavesaved
7. 07
   

   newGPU Mode Lecture 80: How FlashAttention 4 Works [video]www.youtube.com

   ▲ 4matt_d401ddiscusssavesaved
8. 08
   

   newMaking FlashAttention-4 faster for inferencemodal.com

   ▲ 3birdculture147ddiscusssavesaved
9. 09
   

   new7 days of autonomous agent search outperformed FlashAttention-4 and CUDNNtwitter.com

   ▲ 3antinucleon225ddiscusssavesaved
10. 10
    

    newFlash Attention 4, running near matmul speeds [pdf]github.com

    ▲ 3verdverm246ddiscusssavesaved
11. 11
    

    newExploring FlashAttention-3/4 optimizations on RTX GPUsriftstack.ai

    ▲ 2gkapur101d💬 1savesaved
12. 12
    

    newFlashAttention-4: Algorithm and Kernel Pipeliningresearch.colfax-intl.com

    ▲ 2skidrow127ddiscusssavesaved
13. 13
    

    newFlashAttention-4: Algorithm and Kernel Pipelining Co-Designresearch.colfax-intl.com

    ▲ 2skidrow140ddiscusssavesaved
14. 14
    

    newMaking FlashAttention-4 faster for inferencemodal.com

    ▲ 2birdculture144ddiscusssavesaved
15. 15
    

    newMaking FlashAttention-4 faster for inferencemodal.com

    ▲ 2matt_d148ddiscusssavesaved
16. 16
    

    newFlashAttention-4research.colfax-intl.com

    ▲ 2maralom246ddiscusssavesaved
17. 17
    

    newFlashAttention-4: Algorithm and Kernel Pipelining Co-Designresearch.colfax-intl.com

    ▲ 1skidrow124ddiscusssavesaved
18. 18
    

    newFlashAttention-4: Algorithm and Kernel Pipelining Co-Designresearch.colfax-intl.com

    ▲ 1skidrow137ddiscusssavesaved
19. 19
    

    newFlash Attention 4www.together.ai

    ▲ 1zagwdt246ddiscusssavesaved
20. 20
    

    newLaunch HN: Magnitude (YC S25) – Self-optimizing inference engine for agentsgithub.com

    ▲ 194anerli37d💬 96savesaved
21. 21
    

    newShow HN: Trained a 12M transformer on an ML framework we built from scratchgithub.com

    ▲ 2caliandbust202d💬 2savesaved
22. 22
    

    newShow HN: Cactus Hybrid: We taught Gemma 4 to know when it's wronggithub.com

    ▲ 191HenryNdubuaku107d💬 44savesaved
23. 23
    

    newShow HN: Needle2: 14MB agentic LLM for phones, wearables, smart home and robotscactuscompute.com

    ▲ 537HenryNdubuaku88d💬 185savesaved
24. 24
    

    newShow HN: Run TRELLIS.2 Image-to-3D generation natively on Apple Silicongithub.com

    ▲ 202shivampkumar201d💬 40savesaved
25. 25
    

    newDeepSeek V4 is out. the best open-source on coding. here's the breakdown

    ▲ 5Alisaqqt197d💬 1savesaved
26. 26
    

    newShow HN: I vibe-coded some unusual transformer modelsgithub.com

    ▲ 1killerstorm549d💬 1savesaved
27. 27
    

    newShow HN: ThinkTotem – turn boring books into engaging conversationsthinktotem.com

    ▲ 1ccarnino499ddiscusssavesaved

· edit