Muse Glimmer 30B

Meta Superintelligence Labs · meta/muse-glimmer-30b

ProviderMeta Superintelligence Labs
Familymuse
Typevlm
Statusactive
Released2026-08-10
Updated2026-08-11
Parameters29.8B
Open weightsyes

Lineage

RelationshipModel
Is a distillation ofMuse Spark

Capabilities

Chain Of Thought · tier-2 Function Calling · tier-2 Multi Step · tier-2 Multilingual · tier-2 Think Budget Control · tier-2 Tool Error Recovery · tier-2

What it fits on

Ordered by predicted speed. Bandwidth sets decode rate; memory decides whether it runs at all.

Every figure here is computed, not measured. Fit is weights at each quantisation against device memory, with a 25% allowance for the KV cache, activations and the OS. Decode rate is the memory-bandwidth roofline at 70% efficiency. Nobody has run this model on these devices.
DeviceMemoryBandwidthBest qualityWeightstok/sFastest
Cerebras WSE-344.0 GB21000000.0 GB/sfp829.78 GB~493675.8~987351.6 q4
NVIDIA GB200 Grace Blackwell Superchip372.0 GB16000.0 GB/sbf1659.55 GB~188.1~752.3 q4
AMD Instinct MI355X288.0 GB8000.0 GB/sbf1659.55 GB~94.0~376.1 q4
NVIDIA B200180.0 GB7700.0 GB/sbf1659.55 GB~90.5~362.0 q4
Google TPU7x (Ironwood)192.0 GB7380.0 GB/sbf1659.55 GB~86.7~347.0 q4
AMD Instinct MI325X256.0 GB6000.0 GB/sbf1659.55 GB~70.5~282.1 q4
AMD Instinct MI300X192.0 GB5300.0 GB/sbf1659.55 GB~62.3~249.2 q4
NVIDIA H200 SXM141.0 GB4800.0 GB/sbf1659.55 GB~56.4~225.7 q4
NVIDIA H100 NVL94.0 GB3900.0 GB/sbf1659.55 GB~45.8~183.4 q4
NVIDIA H100 SXM80.0 GB3350.0 GB/sbf1659.55 GB~39.4~157.5 q4
AMD Instinct MI250X128.0 GB3200.0 GB/sbf1659.55 GB~37.6~150.5 q4
Google TPU v5p95.0 GB2765.0 GB/sbf1659.55 GB~32.5~130.0 q4
NVIDIA A100 80GB SXM80.0 GB2039.0 GB/sbf1659.55 GB~24.0~95.9 q4
NVIDIA H100 PCIe80.0 GB2000.0 GB/sbf1659.55 GB~23.5~94.0 q4
NVIDIA GeForce RTX 509032.0 GB1792.0 GB/sq622.33 GB~56.2~84.3 q4
Google TPU v6e (Trillium)32.0 GB1638.0 GB/sq622.33 GB~51.3~77.0 q4
AMD Instinct MI21064.0 GB1600.0 GB/sfp829.78 GB~37.6~75.2 q4
NVIDIA A100 40GB SXM40.0 GB1555.0 GB/sfp829.78 GB~36.6~73.1 q4
Google TPU v432.0 GB1200.0 GB/sq622.33 GB~37.6~56.4 q4
NVIDIA GeForce RTX 409024.0 GB1008.0 GB/sq414.89 GB~47.4~47.4 q4
AMD Radeon RX 7900 XTX24.0 GB960.0 GB/sq414.89 GB~45.1~45.1 q4
NVIDIA GeForce RTX 309024.0 GB936.0 GB/sq414.89 GB~44.0~44.0 q4
NVIDIA L40S48.0 GB864.0 GB/sfp829.78 GB~20.3~40.6 q4
Apple M3 Ultra512.0 GB819.0 GB/sbf1659.55 GB~9.6~38.5 q4
AMD Radeon RX 7900 XT20.0 GB800.0 GB/sq414.89 GB~37.6~37.6 q4
Apple M2 Ultra192.0 GB800.0 GB/sbf1659.55 GB~9.4~37.6 q4
Apple M4 Max128.0 GB546.0 GB/sbf1659.55 GB~6.4~25.7 q4
Apple M1 Max64.0 GB400.0 GB/sfp829.78 GB~9.4~18.8 q4
Apple M2 Max96.0 GB400.0 GB/sbf1659.55 GB~4.7~18.8 q4
Apple M3 Max128.0 GB400.0 GB/sbf1659.55 GB~4.7~18.8 q4
NVIDIA L424.0 GB300.0 GB/sq414.89 GB~14.1~14.1 q4
Apple M4 Pro64.0 GB273.0 GB/sfp829.78 GB~6.4~12.8 q4
Apple M432.0 GB120.0 GB/sq622.33 GB~3.8~5.6 q4

Where it runs

Taken from the card's availability section.

PlatformId on that platform
Fireworks AIaccounts/fireworks/models/muse-glimmer-30b
Hugging Facemeta-models/Muse-Glimmer-30B
MLX Communitymlx-community/Muse-Glimmer-30B-OptiQ-4bit
OpenRoutermeta/muse-glimmer-30b
Together AImeta-models/Muse-Glimmer-30B

Verified benchmark evidence

Each row was checked against its source by a reviewer, and carries the model identifier as actually evaluated — which is not always the same as this card's.
BenchmarkEvaluated asScoreEvidence dateSource kind
arena_elo_style_controlmuse-glimmer1427.34 elo2026-09-10 evaluatedindependent evaluatorsource

Reported benchmark scores

This card reports no benchmark scores yet.

BenchmarkCatalogue standingScore

Data

This card as JSON · See it in the graph · Edit on GitHub