Vllm is a topic tracked in our intelligence system with 5 linked articles.
SambaNova raises $1B at an $11B valuation in a first close of its Series F, five months after a mega round, with JPMorgan as a customer and Intel partnership, signaling continued demand for AI inference hardware and potential exit paths.
NVIDIA Cosmos 3 is an open-source physical AI foundation model family (Nano 16B and Super 64B) with a unified Reasoner-Generator MoT architecture, open datasets, post-training workflows, and production-ready NIM microservices for deployment.
Liquid AI unveils LFM2.5-8B-A1B, an 8B parameter MoE edge model with 128K context, 38T pretraining, expanded tokenizer, and strong on-device benchmarking and tool-calling capabilities.
Critical Starlette vulnerability CVE-2026-48710 (BadHost) enables Host header-based path auth bypass in Starlette <1.0.1, affecting thousands of AI infra apps; fix by upgrading to 1.0.1+, adopting endpoint-based security, and placing a reverse proxy in front of ASGI servers.
ZAYA1-8B is a sub-1B active-parameter open-source MoE model (8.4B total) trained entirely on AMD hardware, achieving competitive math/coding benchmarks and highlighting an AMD-focused pathway with open weights and proprietary inference tech.
Open-source test runner for Agent Skills to empirically validate SKILL efficacy via with_skill vs baseline and judge scoring, producing artifacts and HTML reports.
Subscribe for real-time topic updates and unlimited access to our intelligence platform.