llamaperf
Calculator
GPUs
Engines
Guides
Blog
Community
MCP
Advertise
Submit
Sign in
Submit
llamaperf
/
GPU
/
report
Muse 30B Glimmer
on Unknown GPU · ExLlamaV3 · 100,000 ctx
Tone: positive
Sep 8, 2026
Share
X
Reddit
LinkedIn
Bluesky
Copy
link
Throughput
30.0 t/s gen
Quant
EXL3-SC 3.00bpw H4 (EXL3)
KV cache
Q8
VRAM reported
12 GB
Summary
User reports Qwen 3.8 27B at SC2.20bpw H3. User prefers Unsloth UD_Q4_K_XL for coding.
View source ↗
Feedback