vLLM Releases Deployment Guide for Muse-Glimmer-30B Local Multimodal Agent
vllm_project · x · 2026-08-10
vLLM Recipes has published a deployment guide for the Muse-Glimmer-30B model. This is a dense 29.6B parameter vision-language model featuring a ViT-G/14 perception encoder and 128K context, designed for local agentic tasks on consumer hardware.
A unique aspect of the model is its output format: it avoids JSON tool calls and <think> tags for reasoning. Instead, it uses channel-scoped message sequences and XML-style ATEM tool calls. Deployment requires the dedicated museglimmer tool-call and reasoning parsers. Multiple weight variants are available, including BF16, NVFP4 quantization, and a draft head for speculative decoding.
Related event: Meta Open-Sources 30B Muse Glimmer for On-Device Agents(108 posts)→
More from Infra
- Amazon Proposes Per-Request Validated Evidence Caching for Production RAG — _reachsumit · 2026-10-16
- LinkedIn's TransX report: sequence transduction for recsys, +6.02% CTR in A/B tests — _reachsumit · 2026-10-16
- llama.cpp PR boosts CPU inference speed with merged AVX2 tiled accumulators — jacek2023 · 2026-10-16
- Zero's next-gen IVM engine runs in a single DO, live-querying a 100GiB R2 database — threepointone · 2026-10-16
- Meme: driving to the chip factory in 2028 captures AI's compute anxiety — beffjezos · 2026-10-16
- One volunteer maintains the time zone file behind 4 billion phones — danbri · 2026-10-16