Flint-40B-A2.5B · Q4_K_M
This page aggregates 1 real-world runs of Flint-40B-A2.5B (Q4_K_M) on NVIDIA H20 with vLLM, contributed by 1 independent source platforms; metrics are averages of published measurements.
94.37 tok/s
Decode
Decode speed
330.77 tok/s
Prefill
Prefill speed
1.13 s
TTFT
Time to first token
70.63 GB
VRAM
VRAM usage
1
Measured runs
1
Independent sources
Other
Source platforms
14 days ago
Last verified
Performance
- vLLM · Q4_K_M (current)Decode 94.37 · Prefill 330.77 ·
- vLLM · Q6_K Decode 79.46 · Prefill 372.23 ·
- Ollama · Q3_K_S Decode 104.76 · Prefill 539.82 ·
Core figures
- Decode (avg)
- 94.37 tok/s
- Prefill (avg)
- 330.77 tok/s
- TTFT (avg)
- 1.13 s
- VRAM (avg)
- 70.63 GB
- MTP acceptance rate
- —
- TTFB
- 0.75 GB
- Power draw
- 318 W
Configuration
Member-level fields are taken from the most recent run
- Model
- Flint-40B-A2.5B
- Quantization
- Q4_K_M
- Framework
- vLLM
- Version
- v0.27
- Context length
- 32768 tokens
- Batch size
- 8
- 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 H20
- Nominal VRAM
- 97 GB
- Measured VRAM (avg)
- 70.63 GB
- OS
- Windows 11 24H2
- Driver
- 570.86
- CUDA
- 12.6
- Power draw
- 318 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
v0.27 · Windows 11 24H2 · CUDA 12.6 · 32768 ctx
94.37 tok/s
Decode
330.77 tok/s
Prefill
1.13 s
TTFT
70.63 GB
VRAM
—
MTP
318 W
Power draw
L2 batch seed 2026-10-22: vLLM on hw_id=56 running model_id=92 at Q4_K_M