Sam's News — ai — 2026-08-13¶

AI¶

7 Meta releases Muse Glimmer 30B open-weight language model¶

Meta releases Muse Glimmer, a 30-billion-parameter open-source LLM under Apache 2.0 license designed for on-device AI agent deployment.

Sources: venturebeat.com Web Search Update to: Meta releases Muse Glimmer 30B open-weight model for local AI deployment

6 Liquid AI releases LFM2.5-VL-3B vision-language model for on-device use¶

Liquid AI has released a 3-billion-parameter vision-language model capable of screen reading, object grounding, and tool calling entirely on-device.

Sources: MarkTechPost RSS, marktechpost.com Web Search

5.5 DataGrout platform enables enterprises to control AI usage, governance, and LLM costs¶

DataGrout offers a platform for enterprises to manage AI usage, establish governance policies, and control large language model expenses.

Sources: Help Net Security RSS

Security¶

6.5 Attackers hijack company AI systems to run up unauthorized bills¶

Cyber attackers are compromising corporate AI accounts and exploiting them to generate massive computational bills on company credit.

Sources: Business Standard RSS

AI Models¶

6 SpaceXAI Releases Grok 4.6 with Advanced Reasoning¶

SpaceXAI released Grok 4.6, an LLM that the company claims outperforms Anthropic's Claude in some reasoning tasks.

Sources: SiliconANGLE RSS, siliconangle.com Web Search

Hardware¶

5.5 Packet-based NPUs balance compute and memory demands in LLM and automotive workloads¶

Analysis of neural processing units examines how packet-based architectures address the shift from compute-bound CNN tasks to memory-intensive edge and automotive AI workloads.

Sources: Semiconductor Engineering RSS

AI Tools¶

4 SelectHub Launches DataGrout AI Governance Platform¶

SelectHub introduced DataGrout, a research lab platform designed to unify enterprise agent deployments and reduce LLM operational costs.

Sources: GlobeNewswire RSS, globenewswire.com Web Search

3 User Replaces Perplexity with Local LLM for Offline Search¶

A user switched from Perplexity to a locally-deployed large language model on Android for instant offline search capabilities.

Sources: Android Police RSS, androidpolice.com Web Search