Release v1.15.0 adds a speed parameter and usage.speed to Messages, adds between_tools to model thinking types, and returns a Console organization's usage-tier spend cap as a read-only limit; docs, tests and CI tweaks included.
Why it matters
Developers can set and read message speed and use between_tools when checking model capabilities; Console organizations can read their usage-tier spend cap to view spend limits and billing visibility.
NVIDIA and AI4Bharat announce Nemotron-pretraining-Indic-v1 with approximately 2.8 trillion tokens across 22 Indian languages and an open, GPU-accelerated data-generation pipeline for translation, quality assessment, and repair.
Why it matters
Developers can access a traceable synthetic corpus and a reusable GPU pipeline to generate, audit, and extend Indic training data for model development in Indian languages.
Release v0.7.0 adds LiliCorr and GLM5-Next MTP speculative decoding, a host-memory GPU MoE cache (--moe-cache-mib, up to 2.2x decode), many new models, server API changes and ggml v0.27.0.
Why it matters
Developers can reduce VRAM use for embedding workloads up to 10x.
Hugging Face