Meta FAIR that updates centroids with exponential moving averages. How does e-means work? Every k-means variant keeps a sum S ...

Anthropic’s Claude Fable 5.2 and Mythos 5.2 score 49.1% on FrontierCode 2.0, and cost $8/$40 per million tokens.

Reason Machines has open-sourced Aspen, a self-hosted Git server for AI agents. It follows OpenAI’s Pando design and sustained 567 pushes per second on one 8-core server in the company’s own test.

Reason Machines has open-sourced Aspen, a self-hosted Git server for AI agents. It follows OpenAI’s Pando design and sustained 567 pushes per second on one 8-core server in the company’s own test.

Meta Muse Ember is a small multimodal agent model scoring 36 on Artificial Analysis at $0.14 per task. Released today ...

Meta FAIR’s e-means is a mini-batch k-means replacement shipped in FAISS 1.16.0. It reaches Lloyd’s 25-epoch loss in 1.4 to 2.6 epochs, with 10-45x GPU speedups and no extra hyperparameters.

Nous Research’s Arete-27B fine-tunes Qwen3.8-27B with 52.8M tokens, cutting propaganda answers from 22% to 0.9% on 803 ...

Google's Elephant-Goldfish Model plugin makes coding agents draft specs, then runs 5 zero-context review stages before any ...

Baidu’s ERNIE Evolve automated research system beat public bests on NanoChat and NanoGPT Speedrun, lifting ERNIE-5.5-Flash ...

Baidu has released ERNIE 5.5 Flash Base, the raw pre-training checkpoint of its new sparse Mixture-of-Experts language model.

Google released Gemini Omni 1.2 Flash, a video generation model with native 15-second clips at $0.07 per second.

NVIDIA shows 3 ways to speed up MoE inference on DGX Spark, lifting Ling-3.0-flash GSM8K decode throughput 42.5%.