Views
No views yet
TQ3_4S tensor type. It requires
turbo-tan/llama.cpp-tq3.
Stock llama.cpp builds without TurboQuant support cannot load it.| File | Size | Purpose |
|---|---|---|
Muse-Glimmer-30B-TQ3_4S.gguf | 13.78 GiB | Main model (4.25 bpw) |
mmproj-Muse-Glimmer-30B-Q8_0.gguf | 2.0 GiB | Vision projection (image input) |
unsloth/Muse-Glimmer-30B-GGUF (from meta-models/Muse-Glimmer-30B, Apache-2.0)TQ3_4S (four-scale turbo quant) with out6k recipe —
output and embedding tensors preserved at q6_K precision1./build/bin/llama-server \
2 -m Muse-Glimmer-30B-TQ3_4S.gguf \
3 --host 127.0.0.1 --port 8080 \
4 -c 32768 -np 1 -ngl 99 -fa on \
5 --reasoning-format deepseek --jinja1./build/bin/llama-server \
2 -m Muse-Glimmer-30B-TQ3_4S.gguf \
3 --mmproj mmproj-Muse-Glimmer-30B-Q8_0.gguf \
4 --host 127.0.0.1 --port 8080 \
5 -c 32768 -np 1 -ngl 99 -fa on \
6 --reasoning-format deepseek --jinja1# separate drafter model required
2--spec-type draft-dflash -md <drafter>.gguf --spec-draft-n-max 3turbo-tan/llama.cpp-tq3 build f755f1ac1,
thinking ON, temperature 0.
| Benchmark | pass@1 |
|---|---|
| HumanEval | 93.3 |
| HumanEval+ | 89.0 |
| MBPP | 89.7 |
| MBPP+ | 74.6 |
| Benchmark | Result |
|---|---|
| Hard86 | 74/86 (86.0%) |
| Suite | Score | Pass rate |
|---|---|---|
| instructfollow | 96.7 | 14/15 |
| coding | 87.5 | 10/12 |
| dataextract | 82.8 | 9/15 |
| reasonmath | 80.0 | 12/15 |
| toolcall | 80.0 | 11/15 |
| speed | 70.8 | 9/9 |
| Config | Result |
|---|---|
| llama-bench pp2048 | 1,155 tok/s |
| llama-bench tg128 | 43.3 tok/s |
| Decode, 8K context, no drafter | 44.6 tok/s |
| Decode, 8K context, DFlash drafter (n_max=3) | 53.7 tok/s (+20%) |
--reasoning off): content exactly ok ✅block_size=16, mask_token_id=201818, n_extract=5