- Hardware
- 1× NVIDIA GeForce RTX 5090
- Draft model
- meta-models/Muse-Glimmer-30B-GGUF
- Workload
- 1024 → 256
- Runs
- 8 eligible
Efficient frontierOther eligible run
Runs
All published runs with the same model build, accelerator setup, and request shape. Benchmark methodology remains attached to each run.
| Runtime | Material parameters | Test load | Req/s | Output tok/s | p95 TTFT | p95 TPOT | p95 E2E | Failures | Run |
|---|---|---|---|---|---|---|---|---|---|
llama.cpp b10689Speculative · Q4_K_M | speculative.type=draft-dflash speculative.n_max=15 ubatch_size=512 context_size=147456 max_num_seqs=16 flash_attention=on weight_artifact=Muse-Glimmer-30B-KQuant-Dynamic-Q4_K_XL.gguf max_model_len_per_slot=9216 max_num_batched_tokens=16384 | c=8 | 537.9 | 2,376.9 ms | 14.07 ms | 4,301 ms | 0 | run_3ipFIwLFOwrmuTaV | |
llama.cpp b10689Speculative · Q4_K_M | speculative.type=draft-dflash speculative.n_max=15 ubatch_size=512 context_size=147456 max_num_seqs=16 flash_attention=on weight_artifact=Muse-Glimmer-30B-KQuant-Dynamic-Q4_K_XL.gguf max_model_len_per_slot=9216 max_num_batched_tokens=16384 | c=1 | 374.4 | 312.7 ms | 1.56 ms | 705.8 ms | 0 | run_0Ngv9Ngq5iFjJAtZ | |
llama.cpp b10689Speculative · Q4_K_M | speculative.type=draft-dflash speculative.n_max=15 ubatch_size=512 context_size=147456 max_num_seqs=16 flash_attention=on weight_artifact=Muse-Glimmer-30B-KQuant-Dynamic-Q4_K_XL.gguf max_model_len_per_slot=9216 max_num_batched_tokens=16384 | c=1 | 374 | 308.9 ms | 1.55 ms | 701.6 ms | 0 | run_OZ046C2bQECSXE-n | |
llama.cpp b10689Speculative · Q4_K_M | speculative.type=draft-dflash speculative.n_max=15 ubatch_size=512 context_size=147456 max_num_seqs=16 flash_attention=on weight_artifact=Muse-Glimmer-30B-KQuant-Dynamic-Q4_K_XL.gguf max_model_len_per_slot=9216 max_num_batched_tokens=16384 | c=1 | 375.8 | 309.6 ms | 1.54 ms | 699.3 ms | 0 | run_-pmSHKJRhRIryo1f | |
llama.cpp b10689Speculative · Q4_K_M | speculative.type=draft-dflash speculative.n_max=15 ubatch_size=512 context_size=147456 max_num_seqs=16 flash_attention=on weight_artifact=Muse-Glimmer-30B-KQuant-Dynamic-Q4_K_XL.gguf max_model_len_per_slot=9216 max_num_batched_tokens=16384 | c=8 | 537.8 | 2,408.9 ms | 13.88 ms | 4,354 ms | 0 | run_6w4n09STcg0arWg9 | |
llama.cpp b10689Speculative · Q4_K_M | speculative.type=draft-dflash speculative.n_max=15 ubatch_size=512 context_size=147456 max_num_seqs=16 flash_attention=on weight_artifact=Muse-Glimmer-30B-KQuant-Dynamic-Q4_K_XL.gguf max_model_len_per_slot=9216 max_num_batched_tokens=16384 | c=1 | 372.7 | 309.3 ms | 1.55 ms | 706.8 ms | 0 | run_-nj5RzOKCXMJ2Viu | |
llama.cpp b10689Speculative · Q4_K_M | speculative.type=draft-dflash speculative.n_max=15 ubatch_size=512 context_size=147456 max_num_seqs=16 flash_attention=on weight_artifact=Muse-Glimmer-30B-KQuant-Dynamic-Q4_K_XL.gguf max_model_len_per_slot=9216 max_num_batched_tokens=16384 | c=1 | 372.7 | 309.6 ms | 1.55 ms | 703.3 ms | 0 | run_SlvPYaK_-LTNjqxw | |
llama.cpp b10689Speculative · Q4_K_M | speculative.type=draft-dflash speculative.n_max=15 ubatch_size=512 context_size=147456 max_num_seqs=16 flash_attention=on weight_artifact=Muse-Glimmer-30B-KQuant-Dynamic-Q4_K_XL.gguf max_model_len_per_slot=9216 max_num_batched_tokens=16384 | c=1 | 372.9 | 312.8 ms | 1.54 ms | 703 ms | 0 | run_7ZuCv8PHg6xQts0P |
- Runtime
- llama.cpp b10689Speculative · Q4_K_M draft
- Material parameters
speculative.type=draft-dflash speculative.n_max=15 ubatch_size=512 context_size=147456 max_num_seqs=16 flash_attention=on weight_artifact=Muse-Glimmer-30B-KQuant-Dynamic-Q4_K_XL.gguf max_model_len_per_slot=9216 max_num_batched_tokens=16384- Test load
- c=8
- Req/s
- Output tok/s
- 537.9
- p95 TTFT / TPOT
- 2,376.9 ms / 14.07 ms
- Runtime
- llama.cpp b10689Speculative · Q4_K_M draft
- Material parameters
speculative.type=draft-dflash speculative.n_max=15 ubatch_size=512 context_size=147456 max_num_seqs=16 flash_attention=on weight_artifact=Muse-Glimmer-30B-KQuant-Dynamic-Q4_K_XL.gguf max_model_len_per_slot=9216 max_num_batched_tokens=16384- Test load
- c=1
- Req/s
- Output tok/s
- 374.4
- p95 TTFT / TPOT
- 312.7 ms / 1.56 ms
- Runtime
- llama.cpp b10689Speculative · Q4_K_M draft
- Material parameters
speculative.type=draft-dflash speculative.n_max=15 ubatch_size=512 context_size=147456 max_num_seqs=16 flash_attention=on weight_artifact=Muse-Glimmer-30B-KQuant-Dynamic-Q4_K_XL.gguf max_model_len_per_slot=9216 max_num_batched_tokens=16384- Test load
- c=1
- Req/s
- Output tok/s
- 374
- p95 TTFT / TPOT
- 308.9 ms / 1.55 ms
- Runtime
- llama.cpp b10689Speculative · Q4_K_M draft
- Material parameters
speculative.type=draft-dflash speculative.n_max=15 ubatch_size=512 context_size=147456 max_num_seqs=16 flash_attention=on weight_artifact=Muse-Glimmer-30B-KQuant-Dynamic-Q4_K_XL.gguf max_model_len_per_slot=9216 max_num_batched_tokens=16384- Test load
- c=1
- Req/s
- Output tok/s
- 375.8
- p95 TTFT / TPOT
- 309.6 ms / 1.54 ms
- Runtime
- llama.cpp b10689Speculative · Q4_K_M draft
- Material parameters
speculative.type=draft-dflash speculative.n_max=15 ubatch_size=512 context_size=147456 max_num_seqs=16 flash_attention=on weight_artifact=Muse-Glimmer-30B-KQuant-Dynamic-Q4_K_XL.gguf max_model_len_per_slot=9216 max_num_batched_tokens=16384- Test load
- c=8
- Req/s
- Output tok/s
- 537.8
- p95 TTFT / TPOT
- 2,408.9 ms / 13.88 ms
- Runtime
- llama.cpp b10689Speculative · Q4_K_M draft
- Material parameters
speculative.type=draft-dflash speculative.n_max=15 ubatch_size=512 context_size=147456 max_num_seqs=16 flash_attention=on weight_artifact=Muse-Glimmer-30B-KQuant-Dynamic-Q4_K_XL.gguf max_model_len_per_slot=9216 max_num_batched_tokens=16384- Test load
- c=1
- Req/s
- Output tok/s
- 372.7
- p95 TTFT / TPOT
- 309.3 ms / 1.55 ms
- Runtime
- llama.cpp b10689Speculative · Q4_K_M draft
- Material parameters
speculative.type=draft-dflash speculative.n_max=15 ubatch_size=512 context_size=147456 max_num_seqs=16 flash_attention=on weight_artifact=Muse-Glimmer-30B-KQuant-Dynamic-Q4_K_XL.gguf max_model_len_per_slot=9216 max_num_batched_tokens=16384- Test load
- c=1
- Req/s
- Output tok/s
- 372.7
- p95 TTFT / TPOT
- 309.6 ms / 1.55 ms
- Runtime
- llama.cpp b10689Speculative · Q4_K_M draft
- Material parameters
speculative.type=draft-dflash speculative.n_max=15 ubatch_size=512 context_size=147456 max_num_seqs=16 flash_attention=on weight_artifact=Muse-Glimmer-30B-KQuant-Dynamic-Q4_K_XL.gguf max_model_len_per_slot=9216 max_num_batched_tokens=16384- Test load
- c=1
- Req/s
- Output tok/s
- 372.9
- p95 TTFT / TPOT
- 312.8 ms / 1.54 ms