Hackernews posts about MoE

1. Show HN: Maple-Preview – Ternary 20B MoE running at 120 tok/s on a iPhone (deepgrove.ai)
   173 points by edwardbzhang 16 days ago | 53 comments
2. Show HN: Lumabri – Run Moe Models on a P2P Swarm with Colibri (github.com)
   48 points by vforno 7 days ago | 19 comments
3. Benchmarking Qwen 3.6 35B MoE (3B active) on an RTX 3090 (www.gilesthomas.com)
   16 points by gpjt 27 days ago | discuss
4. Ask HN: HotPin – lossless 120B MoE inference on 24GB RAM (CPU, 50 loc)
   14 points by LozzKappa 26 days ago | discuss
5. Mixture-of-Kittens: our open-source MoE megakernel for NVL72s (cursor.com)
   12 points by mfiguiere 16 days ago | discuss
6. Show HN: RunNburn – Run a 295B Moe from a 98GB GGUF on a 64GB RAM Desktop (github.com)
   11 points by coderredlab 22 days ago | discuss
7. AntGroup releases Ling 3.0 Flash MoE model (twitter.com)
   4 points by potus_kushner 22 days ago | discuss
8. Show HN: Hardware-software co-design for MoE models to bypass NCCL bottlenecks (github.com)
   3 points by PJHkorea 16 days ago | 1 comments
9. Show HN: Ullis – Local Ternary Moe-Kan Training and Inference Engine in Rust (github.com)
   3 points by vlad_kalinkin about 13 hours ago | discuss
10. Ornith 1.5 35B-A3B: SOTA 35B MoE model (huggingface.co)
    3 points by nateb2022 1 day ago | discuss
11. DeepSeek V4 Pro goes GA: 1.6T MoE flagship exits preview (ainexusdaily.vercel.app)
    3 points by qrify_app 6 days ago | discuss
12. Mixture-of-Kittens: An MoE training megakernel for NVL72 (twitter.com)
    3 points by OsamaJaber 16 days ago | discuss
13. Instella-Moe: An Open Mixture-of-Experts Language Model (rocm.blogs.amd.com)
    3 points by ilreb 27 days ago | discuss
14. Mixture-of-Kittens: our open-source MoE megakernel for NVL72s (cursor.com)
    2 points by gmays about 10 hours ago | discuss
15. Imprint – Fine-tune MoE LLMs bigger than your RAM (github.com)
    2 points by pyeAI 9 days ago | discuss
16. MoE routing is just branch prediction (ssenthilnathan3.github.io)
    2 points by nathaah3 11 days ago | discuss
17. Mixture-of-Kittens: Cursor's open-source MoE megakernel for NVL72s (cursor.com)
    2 points by alexhandru 16 days ago | discuss
18. Show HN: SkewAdam – A tiered optimizer that cuts MoE state memory by 97% (github.com)
    2 points by anonnumaan 30 days ago | discuss
19. Train-infer mismatch for Open-weight MoE RL in Open-source code (kiddyboots216.github.io)
    1 points by kiddyboots216 3 days ago | discuss
20. Solving Moe Load Imbalance in LLM Training via Optimal Transport (arxiv.org)
    1 points by nullnonenilNULL 4 days ago | discuss
21. Mixture of Experts (Moe): How Transformers Scale Without Activating Everything (chizkidd.github.io)
    1 points by ibobev 10 days ago | discuss
22. The Voyage 4 model family: shared embedding space with MoE architecture (blog.voyageai.com)
    1 points by Topfi 21 days ago | discuss
23. Generate LoRA adapters from Skill.MDs *MoE-models now too (terradev.cloud)
    1 points by Facingsouth 24 days ago | discuss
24. Benchmarking Qwen 3.6 35B MoE (3B active) on an RTX 3090 (www.gilesthomas.com)
    1 points by ibobev 25 days ago | discuss
25. Show HN: I shrank DeepSeek V4 Flash to 57GB and it wrote a compiler on my Mac (huggingface.co)
    21 points by hacklas 4 days ago | 3 comments
26. Compute-optimal is not cluster-optimal (szha.ai)
    13 points by yuxin_tang 7 days ago | discuss
27. Ask HN: Is neuromorphic computing going to replace traditional AI?
    5 points by lennart-rth 27 days ago | 3 comments
28. Log is non-monotonic in PHP and Lua (purplesyringa.moe)
    70 points by ibobev 29 days ago | 55 comments
29. Beyond "Clean Code": Why Your Comments Matter (blog.moertel.com)
    23 points by chmaynard 24 days ago | 10 comments
30. Amazon once again locking me out of my own order data (bsky.app)
    6 points by jmole 29 days ago | 3 comments