Predictions on the Next-Gen Sparse LLM Route
_xjdr · x · 2026-07-17
The author expresses no surprise at a certain TM approach, noting it aligns perfectly with what they've consistently recommended to Western labs.
Core predictions:
- Adopt an architecture like DSV3 shape, but with at least a 4:1 sliding window.
- Train with k2 rollouts, scaling parameters to at least 1T.
- Combine muon and mup, training on GB300.
- This is fundamentally an nmoe route.
They add that Meta should have taken this path from the start with Llama 4 and beyond. Looking ahead, they hope to see larger 3T+ models, higher expert sparsity, KV cache compression, base model releases, frontier RL refinement, and distillation papers. They believe that with higher sparsity, this architecture won't be the main bottleneck in the short term.
Related event: Kimi K3 Debuts Strong, Narrowing the Open-Weight Gap(184 posts)→
More from Models
- Gemini 3.6 Flash lands in Google AI Studio with cheaper output pricing — gaganghotra_ · 2026-07-21
- Jack Clark says OpenAI’s internal-deployment safety notes help the whole frontier community — jackclarkSF · 2026-07-21
- Mindlab Research puts Macaron-V1-Venti on Hugging Face — External_Mood4719 · 2026-07-21
- Google is surfacing Gemini 3.5 Flash-Lite and 3.6 Flash in AI Studio — Expensive_Syrup_6529 · 2026-07-21
- ChatGPT often explains the wall before answering whether it is tilting — Aware-sky-3489 · 2026-07-21
- Grok website traffic rose 38.15% YoY to 736 million Q2 visits — XFreeze · 2026-07-21