Articles

Qwen 3.8, Qwen 3.6 and Muse Glimmer compared

Qwen 3.8's weights landed on schedule, and the 27B walks into a real contest against its own 3.6 sibling and Meta's Muse Glimmer - three dense models sharing one 24 GB shelf.

Aug 16, 2026 · Infersec Team

Muse Glimmer and other great lightweight US models

Meta's Muse Glimmer 30B leads a quiet run of capable US-made models that fit consumer hardware - though Qwen 3.8-Max lands 12 August 2026 and may shuffle the order.

Aug 10, 2026 · Infersec Team

Your own sovereign AI provider with Infersec Self-Hosted

Run the full Infersec platform on your own hardware or private cloud - a sovereign, self-hosted AI provider with no per-request billing.

Aug 6, 2026 · Infersec Team

Qwen 3.8 and other great models to try locally now

Alibaba's Qwen 3.8-Max claims 2.4 trillion parameters and second place behind Claude Fable 5 - and it's just one of six open-weight frontier models that landed in July 2026, the rest already runnable on hardware you own.

Aug 4, 2026 · Infersec Team

Build a dependable agentic platform with Mastra

Mastra gives you agents, workflows, memory and tracing you can actually depend on - but it's only as private as the model behind it. Point it at local weights on your own hardware instead.

Jul 17, 2026 · Infersec Team

Private AI chat for your team, on hardware you own

Give your team a ChatGPT they can actually use at work, running on hardware you own, through an EU-hosted backend, with a desktop chat client they more or less already know how to use.

Jun 29, 2026 · Infersec Team

How to self-host a business-ready Hermes Agent with GBrain

Hermes Agent is the new self-hosted framework that feels built for real use, and here is how to run it with GBrain memory on your own hardware via Infersec.

Jun 25, 2026 · Infersec Team

The open-weight model situation just got ridiculous - and you can run all of it yourself

Anthropic's Claude Fable 5 got pulled three days after launch by a US export order. Meanwhile Gemma 4 with QAT, Liquid's LFM2.5, and Qwen 3.6 quietly closed the open-weight gap - and you can run all of them at scale on hardware you own.

Jun 17, 2026 · Infersec Team

How to Power Opencode With Your Own AI Hardware Using Infersec

Run Opencode with local models on hardware you own. No API keys from big providers, no per-token bills. Self-hosted inference through an OpenAI-compatible API, powered by Infersec.

Jun 3, 2026 · Infersec Team

Become your own AI provider: Infersec Early-Access

Turn hardware you already own into a production-grade AI API. OpenAI and Anthropic compatible, with European hosting, MCP tool calling, and pay-as-you-go pricing.

May 27, 2026 · Infersec Team