SpikingBrain 2.0 Breaks Long‑Sequence and Low‑Power Bottlenecks in Brain‑Inspired LLMs

The Chinese Academy of Sciences unveils SpikingBrain 2.0‑5B, a brain‑inspired large model that uses dual‑space sparse attention and dual activation (FP8 and INT8‑Spiking) to cut training cost by over tenfold, achieve up to 15× speedup on long sequences, and match Qwen‑3 performance while drastically reducing power consumption.

Large Language ModelSparse AttentionSpikingBrain2.0

0 likes · 10 min read

SpikingBrain 2.0 Breaks Long‑Sequence and Low‑Power Bottlenecks in Brain‑Inspired LLMs

Huawei Cloud Developer Alliance

Dec 2, 2025 · Artificial Intelligence

Why Hierarchical Reasoning Model (HRM) Beats Large Models on ARC with Few Samples

The Hierarchical Reasoning Model (HRM) draws on brain-inspired hierarchical processing to achieve state‑of‑the‑art performance on ARC, Sudoku‑Extreme, and Maze‑Hard tasks using under a thousand training examples, while exposing generalization limits on private test sets.

ARC BenchmarkAdaptive Computation TimeHierarchical Reasoning

0 likes · 11 min read

Why Hierarchical Reasoning Model (HRM) Beats Large Models on ARC with Few Samples

HyperAI Super Neural

Sep 30, 2025 · Artificial Intelligence

SpikingBrain-1.0 Achieves 100× Faster Inference with Brain‑Inspired Spiking Architecture

SpikingBrain-1.0, the first domestically‑produced brain‑inspired spiking large model, links spiking neuron dynamics to linear attention, delivering over 100× faster first‑token latency on 4‑million‑token sequences, 23.4% FLOP utilization, 69% sparsity, and a one‑click deployment tutorial on HyperAI.

Large Language ModelSpikingBrain-1.0brain-inspired AI

0 likes · 7 min read

SpikingBrain-1.0 Achieves 100× Faster Inference with Brain‑Inspired Spiking Architecture

SpikingBrain 2.0 Breaks Long‑Sequence and Low‑Power Bottlenecks in Brain‑Inspired LLMs

Why Hierarchical Reasoning Model (HRM) Beats Large Models on ARC with Few Samples

SpikingBrain-1.0 Achieves 100× Faster Inference with Brain‑Inspired Spiking Architecture

SpikingBrain 2.0 Breaks Long‑Sequence and Low‑Power Bottlenecks in Brain‑Inspired LLMs