[models]By ByteBulletin Editor
Reflection AI releases Beam open-weight model
The 501-billion-parameter mixture-of-experts model claims to match Chinese rivals on reasoning benchmarks while using 3-4x less inference compute.
[tag]
#mixture-of-experts
7 stories
[research]By ByteBulletin Editor
ExFold-MoE: A New Mixture-of-Experts Approach for Efficient and Accurate Protein Structure Prediction
A novel mixture-of-experts architecture promises to make protein structure prediction both faster and more accurate, with potential implications for AI-driven drug discovery.
[research]By ByteBulletin Editor
Entropymoe: A New Entropy-Based MoE Framework Promises More Efficient Sparse Model Training
Researchers propose an entropy-driven routing mechanism for mixture-of-experts models that could reduce compute overhead while maintaining accuracy.
[launches]By ByteBulletin Editor
Lumabri Lets You Run Huge Mixture-of-Experts Models Across a Swarm of Peer Machines
A new pure-C, dependency-free system pools CPU and disk from any machines on a network to serve massive MoE models, with byte-identical output and no upfront downloads.
[models]By ByteBulletin Editor
Escha Labs ships 2-bit Qwen3.6-35B-A3B build that runs on a 24GB GPU
A 12.3GB MoE quant packs a 35B model onto consumer cards, with quality within noise of FP8 on most benchmarks.
[research]By ByteBulletin Editor
ExpertCache lets a 120B Mixture-of-Experts model run on a 16GB Mac — slowly, but surprisingly intact
Amos Labs' experimental Metal runtime proves oversized sparse MoE checkpoints can fit in constrained Apple Silicon memory without sacrificing functional capability, even if interactive speeds remain out of reach.
[research]By ByteBulletin Editor
Sticky Routing MoE: Reducing Expert Switching for More Efficient Inference
New research proposes sticky routing for mixture-of-experts models, keeping tokens on the same expert across layers to cut communication overhead and improve throughput.
