DeepSeek V4.1 Leak: 552B New Architecture, HBM Needs Cut 3.9x, Huge Benchmark Gains
teortaxesTex · x · 2026-09-10
Unofficial benchmarks for DeepSeek V4.1 show a new 552B-parameter Causal-Encoder-Decoder architecture (8B active / 16B output) posting strong numbers: 31.2 on TerminalBench 4.0, 88.1 on CyberGym, 15.3 on ExploitGym, 63.9 on HLE with tools, and 54.8 on Automation-Bench.
On efficiency, HBM requirements are reportedly slashed 3.9x and SSD usage 8x versus V4-Flash, with KV cache efficiency still unmatched six months after V4's arch debuted. The poster estimates >90% inference margins and quips that an American lab with these results would be labeled "recursive self-improvement." Figures are unverified leaks.
Related event: DeepSeek V4.1 and V4.1 Flash Benchmarks and Architecture Allegedly Leak(8 posts)→
More from Models
- vLLM Ships Full Support for DeepSeek-V4.1-Flash's New Architecture — vllm_project · 2026-09-10
- Pro 5x user rage-quits over Codex usage limits, says Plus gets more coding done — DrunkenPionier · 2026-09-10
- DeepSeek V4.1 Flash rumored days away: 522B params, 1/4 KV cache, vision added — R_Duncan · 2026-09-10
- DeepSeek-V4.1-Flash reportedly offers continuous reasoning effort from 1 to 100 — zainhas · 2026-09-10
- DeepSeek multimodal team praised for unusually careful pre-training data work — zephyr_z9 · 2026-09-10
- DeepSeek v4.1 flash evals leak: SOTA on deepswe, competitive on terminal bench — zainhas · 2026-09-10