MiniMax released M3.1 Flash, a 428B-parameter mixture-of-experts model with approximately 23B active parameters and 1M-token context window (MarkTechPost). API pricing is set at $0.15 per 1M input tokens and $0.60 per 1M output tokens, with open weights available October 23 (MarkTechPost). The model targets agentic workloads with extended context support.
NVIDIA open-sourced the Run:ai GPU orchestration platform under Apache 2.0 license, including the KAI Scheduler, CUDA 13.4 GPU fractioning, quota management, and a local Kind development environment (MarkTechPost). Mistral AI released Timestral, a time series forecasting model family ranging from 5M to 209M parameters, with the 209M variant outperforming Google’s TimesFM-3 on zero-shot benchmarks and running on CPU (MarkTechPost).