CLAIM-2026-08-15-02
Meta's own post states that Muse Glimmer at full precision would require over 55 GB of memory, that quantisation to approximately 4-bit shrinks the language model to under 20 GB, and that a K-Quant build fits within a 24 GB or 32 GB envelope alongside working memory and the perception encoder.
Why it matters
It fixes the deployment picture concretely. A 24 GB workstation card is enough, which is why the fine-tuning question stops being theoretical for small teams.
Sources
- metaairesea2026introduci