Hackernews posts about MoE

1. Moebius: 0.2B image inpainting model with 10B-level performance (hustvl.github.io)
   322 points by DSemba 14 days ago | 83 comments
2. LongCat-2.0, a large-scale MoE model with 1.6T total and 48B Active (longcat.chat)
   279 points by benjiro29 7 days ago | 87 comments
3. A 35B MoE on a 16 GB GPU, without the offload tax (www.lucebox.com)
   18 points by GreenGames 28 days ago | discuss
4. Expert Selections in MoE Transformer Models Reveal Almost as Much as Text (arxiv.org)
   5 points by busserweiser 29 days ago | discuss
5. Running a 35B MoE model on a 2017 AMD RX 580 8GB via Vulkan (no ROCm/CUDA) (github.com)
   4 points by aivisionslab 16 days ago | discuss
6. Cohere North Mini Code 30B Moe Apache 2.0 (cohere.com)
   3 points by jeromechoo 27 days ago | 1 comments
7. Porting Moebius 0.2B image inpainting model to run in browser with Claude Code (simonwillison.net)
   3 points by lumpa 14 days ago | discuss
8. Moebius: 0.2B Lightweight Image Inpainting Framework with 10B-Level Performance (hustvl.github.io)
   3 points by ilreb 17 days ago | discuss
9. Moe Estimator – Simulate decode speed with layer-major prefetch hiding (www.agrillo.it)
   2 points by ConteMascetti71 2 days ago | discuss
10. Modular 26.4: SOTA Moe Serving, Model Bringup via Agent Skills, Mojo 1.0 Beta 2 (www.modular.com)
    2 points by tosh 9 days ago | discuss
11. Moebius: Tangled Up in Blueberry (tomlennon.com)
    2 points by rafaepta 14 days ago | discuss
12. Mixture-of-Experts (Moe), Explained: Why "Active Parameters" Decide What Runs (vettedconsumer.com)
    2 points by ermantrout 25 days ago | discuss
13. MoE expert co-activations: Reordering inputs yields easy throughput gains (blog.doubleword.ai)
    2 points by kkm 28 days ago | discuss
14. Efficient and Lossless Moe Diffusion LLM Inference with I/O-Aware Expert Offload (tide-paper.vercel.app)
    1 points by imalomder 27 days ago | 1 comments
15. Sharded Inference of a 229B-Parameter Moe over the Internet at Interactive Speed (twitter.com)
    1 points by MrBuddyCasino 2 days ago | discuss
16. Porting the Moebius 0.2B image model to run in Claude Code on web (simonwillison.net)
    1 points by ddp26 4 days ago | discuss
17. Show HN: Running Gemma-4 26B at 124 tokens/SEC on a CPU, no GPU (apeg.dev)
    10 points by arun-prasath 6 days ago | 1 comments
18. Story of How Im Running an Unlimited $6/Month AI Provider on 4x RTX 3090s
    8 points by yolo-auto 23 days ago | 3 comments
19. I patched llama.cpp to gain 20% prompt processing TPS. Help me make a PR
    6 points by i_am_rocoe 9 days ago | 2 comments
20. LongCat-2.0
    6 points by rika321 1 day ago | discuss
21. Show HN: Levi – run AlphaEvolve on your Claude Code/Codex for dirt cheap (ttanv.github.io)
    5 points by ttanv 28 days ago | discuss
22. As; HN: I was curious why MTP affects PP TPS in llama.cpp. My PoC recovers it?
    4 points by i_am_rocoe 11 days ago | 1 comments
23. The US Government Is Now a Shareholder in 26 Companies (moeonmargin.substack.com)
    128 points by measurablefunc 4 days ago | 116 comments
24. Recovering garbled Bitcoin addresses (2024) (purplesyringa.moe)
    14 points by zX41ZdbW 2 days ago | 2 comments
25. Various projects disappeared from kernel.org hosting overnight (gts.q66.moe)
    10 points by marvinborner 4 days ago | 1 comments
26. Removing Idle Uploads (blog.catbox.moe)
    5 points by Cider9986 4 days ago | 1 comments
27. Using GitHub Actions to Build Linux Kernels – 萌え豚's Blog (moebuta.org)
    5 points by abdelhousni 10 days ago | discuss
28. Show HN: An mkv player that uses WASM to render you videos (parallax.kinosoft.moe)
    4 points by kafkaesqueKino 29 days ago | discuss
29. DeepSeek v4 Releasing Mid July (files.catbox.moe)
    2 points by RIshabh235 7 days ago | 3 comments
30. Suffix BWT vs. cyclic shift BWT, and fast computation (purplesyringa.moe)
    1 points by g0xA52A2A 3 days ago | discuss