Leak: DeepSeek V4.1 reportedly drops V4's HCA, generalizes CSA, removes sparse-attention warmup
zephyr_z9 · x · 2026-09-10
KOL teortaxesTex claims DeepSeek V4.1 strips out V4's workarounds at every level: HCA is ditched, CSA is generalized, no sparse-attention warmup, a simpler single-pass mHC, plus Engram, calling it a 'new Transformer'. He earlier predicted DeepSeek would compress V4's architecture into a simpler set of primitives, arguing DSA isn't the limit and only real architectural exploration crosses deep valleys. Unverified rumor.
Related event: Leaked DeepSeek V4.1 Benchmarks Point to New 552B Architecture(15 posts)→
More from Models
- DeepSeek V4.1 Flash is actually 748B params, safetensors analysis shows — DistanceSolar1449 · 2026-09-10
- Kimi K3 lands on RunPod: 2.8T params, 1M context, $3/$15 per 1M tokens — Kimi_Moonshot · 2026-09-10
- Dev burns 300M tokens on GLM 5.3 in a week and still has quota left — saibharadwaj · 2026-09-10
- Follow-up: a 3T-parameter model may already exist, scaling issues remain the wildcard — teortaxesTex · 2026-09-10
- Speculation: DeepSeek V4.1 Pro could be a 3.1T-param MoE with 2.6TB disk footprint — teortaxesTex · 2026-09-10
- New Book Teaches Beginners to Build and Fine-Tune Their Own GPT-Style SLMs, With Colab Notebooks — Roger_M_Taylor · 2026-09-10