Deep Dive into a Model's Attention and MoE Details
nrehiew_ · x · 2026-07-16
This post breaks down a model's architectural design, focusing on two main areas:
Attention Design
- Uses sliding window, which is somewhat surprising given the team's strong background in linear attention.
- Applies basic 1D convolution on KV and residual.
- No RoPE is used; instead, attention weights are formed by projecting the hidden state and combining it with distance-dependent position bias.
MoE Design
- The model scale is about 975B parameters with 41B active, trained on over 45T multimodal tokens.
- Compared to Dsv3, active parameters increased by roughly 10%, and token count is 3 times larger.
- Features 6 routed experts out of 256 experts, plus 2 shared experts; the author notes that having 2 shared experts is quite rare, as the norm is typically 0 or 1.
Overall, a highly detailed architectural observation post.
Related event: Inkling Performance and Related Architecture Speculation(7 posts)→
More from Models
- MissionReport benchmark: only 3 of 540 AI-generated government reports pass all criteria — Exp_Mark · 2026-10-27
- Local beats cloud: Qwen 3.8 FN handles private student data as well as frontier models — dr_lm · 2026-10-27
- Aidan Gomez: model's CoT said no tool calls right before making them — aidangomez · 2026-10-27
- Nat Lambert: Cohere's model is narrow RL training on internal infra ill-suited for open weights — natolambert · 2026-10-27
- OpenAI teases DOTS is coming to Europe 'soon' — mark_k · 2026-10-27
- Clifford Pickover challenges AI to decipher a message he encoded in Norse runes — pickover · 2026-10-27