PulseAugur
EN
LIVE 16:27:23

GLM-5.3-Flash (321B) runs locally on single RTX 5090 via Project Maya

By PulseAugur Editorial · [1 sources] ·

A technical guide details how to run the large GLM-5.3-Flash (321B) language model locally on a single RTX 5090 GPU with 128 GB of RAM. The process involves using Project Maya, a fork of the Strata runtime, which intelligently manages model components across the GPU, system RAM, and NVMe storage. This approach allows for the execution of a 125 GB EXL3 quant of the model, achieving approximately 25 tokens per second with a 512K-token context window. AI

IMPACT Enables local execution of large models on consumer hardware, lowering barriers for experimentation and development.

RANK_REASON Guide on running a specific LLM locally using a particular software stack.

Read on dev.to — LLM tag →

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →

GLM-5.3-Flash (321B) runs locally on single RTX 5090 via Project Maya

How we ranked this

Signal score
32 / 100
Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket
Tool
Guide on running a specific LLM locally using a particular software stack.
Source corroboration
Single-source cluster
Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics
infra, model release
Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance
High
Clearly on-topic for AI-industry coverage.
Story freshness
Breaking (< 6h)
Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

Full methodology in our editorial standards.

COVERAGE [1]

  1. dev.to — LLM tag TIER_1 English(EN) · Seppe Gadeyne ·

    Run GLM-5.3-Flash (321B) locally with Project Maya and EXL3 on one RTX 5090

    <p><strong>GLM-5.3-Flash is a 321B mixture-of-experts model, and its 3.05 bpw EXL3 quant is 125 GB, too big to fit in my workstation's 128 GB of RAM next to the OS.</strong> It still runs locally here: on one RTX 5090 with 128 GB of RAM, with a 512K-token context, at about 25 tok…

RELATED ENTITIES

RELATED TOPICS