Flint-40B-A2.5B · Q3_K_L
This page aggregates 1 real-world runs of Flint-40B-A2.5B (Q3_K_L) on NVIDIA RTX PRO 6000 Blackwell 96GB with vLLM, contributed by 1 independent source platforms; metrics are averages of published measurements.
327.3 tok/s
Decode
Decode speed
1,746.79 tok/s
Prefill
Prefill speed
0.99 s
TTFT
Time to first token
62.28 GB
VRAM
VRAM usage
1
Measured runs
1
Independent sources
Other
Source platforms
14 days ago
Last verified
Performance
- vLLM · Q3_K_L (current)Decode 327.3 · Prefill 1746.79 ·
- Strata · Q3_K_S Decode 343.48 · Prefill 1507.36 ·
Core figures
- Decode (avg)
- 327.3 tok/s
- Prefill (avg)
- 1,746.79 tok/s
- TTFT (avg)
- 0.99 s
- VRAM (avg)
- 62.28 GB
- MTP acceptance rate
- —
- TTFB
- 2.62 GB
- Power draw
- 389.5 W
Configuration
Member-level fields are taken from the most recent run
- Model
- Flint-40B-A2.5B
- Quantization
- Q3_K_L
- Framework
- vLLM
- Version
- v0.55
- Context length
- 4096 tokens
- Batch size
- 16
- GPU layers
- 999
- Flash Attention
- On
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)
- 62.28 GB
- OS
- Windows 11 24H2
- Driver
- 570.86
- CUDA
- 12.8
- Power draw
- 389.5 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
v0.55 · Windows 11 24H2 · CUDA 12.8 · 4096 ctx
327.3 tok/s
Decode
1,746.79 tok/s
Prefill
0.99 s
TTFT
62.28 GB
VRAM
—
MTP
389.5 W
Power draw
L2 batch seed 2026-10-22: vLLM on hw_id=68 running model_id=92 at Q3_K_L