新加坡初创 Featherless 将接入 MiniMax M3.1 Flash 用于编程

Attention variants like MLA, Mamba-2 hybrids and sparse attention are replacing vanilla transformers as the default choices in frontier open weights, so this roundup is a fast map of which lab bet on what. Practitioners choosing models to self-host can scan parameter counts and mechanisms before committing GPU budget. Since most of these weights are openly released, teams in Singapore and ASEAN can evaluate them locally without waiting for regional API availability.

官方账号Sebastian Raschka@rasbtSource

Accuracy on agentic retrieval benchmarks may already be saturated, so the real design trade-off is token cost versus latency rather than recall. Builders of agentic RAG systems should benchmark indexes on tokens and latency instead of assuming richer structures lift accuracy. The 38% average token reduction with a decision gate keeping latency near 30 seconds offers a concrete configuration to test against dense-retrieval baselines.

DAIR.AI@dair_ai2 sourcesSource

Quantization now puts a 428B open model within reach of a 4× RTX 3090 rig or even Apple Silicon, a threshold few labs have crossed. Local deployment practitioners get ready-made quant formats, expert-offloading recipes and fidelity benchmarks to start from. The featured GGUF and EXL3 builds compete directly with llama.cpp-style workflows for larger but denser open models.

MiniMax_AI@MiniMax_AISource