mammoth-1 · IQ3_XS
This page aggregates 1 real-world runs of mammoth-1 (IQ3_XS) on NVIDIA RTX PRO 6000 Blackwell 96GB with vLLM, contributed by 1 independent source platforms; metrics are averages of published measurements.
228.17 tok/s
Decode
Decode speed
758.2 tok/s
Prefill
Prefill speed
1.08 s
TTFT
Time to first token
37.25 GB
VRAM
VRAM usage
1
Measured runs
1
Independent sources
Other
Source platforms
13 days ago
Last verified
Performance
- vLLM · IQ3_XS (current)Decode 228.17 · Prefill 758.2 ·
- vLLM · Q5_K_M Decode 145.66 · Prefill 731.49 ·
- Ollama · IQ3_XS Decode 214.08 · Prefill 806.7 ·
- Ollama · Q5_K_S Decode 135.94 · Prefill 501.24 ·
Core figures
- Decode (avg)
- 228.17 tok/s
- Prefill (avg)
- 758.2 tok/s
- TTFT (avg)
- 1.08 s
- VRAM (avg)
- 37.25 GB
- MTP acceptance rate
- —
- TTFB
- 1.83 GB
- Power draw
- 328 W
Configuration
Member-level fields are taken from the most recent run
- Model
- mammoth-1
- Quantization
- IQ3_XS
- Framework
- vLLM
- Version
- v0.51
- Context length
- 4096 tokens
- Batch size
- 8
- GPU layers
- 999
- Flash Attention
- Off
Hardware
Nominal and measured figures are shown side by side; whether it runs is the reader's call
- GPU
- NVIDIA RTX PRO 6000 Blackwell 96GB
- Nominal VRAM
- 96 GB
- Measured VRAM (avg)
- 37.25 GB
- OS
- Ubuntu 22.04
- Driver
- 560.94
- CUDA
- 12.4
- Power draw
- 328 W
Get started
Original model resources
Base model resources may not match this quantization. Use the verified deployment weights above when available.
Sources & evidence
1 measured records in total, each traceable to its original source
v0.51 · Ubuntu 22.04 · CUDA 12.4 · 4096 ctx
228.17 tok/s
Decode
758.2 tok/s
Prefill
1.08 s
TTFT
37.25 GB
VRAM
—
MTP
328 W
Power draw
L2 batch seed 2026-10-22: vLLM on hw_id=68 running model_id=94 at IQ3_XS