Back to search results
Kimi-K3 · IQ3_S
NVIDIA RTX 509032GBStrata v0.39L2 Cross-framework verified
This page aggregates 1 real-world runs of Kimi-K3 (IQ3_S) on NVIDIA RTX 5090 with Strata, contributed by 1 independent source platforms; metrics are averages of published measurements.
330.44 tok/s
Decode
Decode speed
1,792.04 tok/s
Prefill
Prefill speed
2.4 s
TTFT
Time to first token
1.05 GB
VRAM
VRAM usage
L2 Cross-framework verified
1
Measured runs
1
Independent sources
Other
Source platforms
Today
Last verified
Performance
- vLLM · Q3_K_L Decode 253.52 · Prefill 1057.1 ·
- vLLM · Q4_K_S Decode 223.8 · Prefill 1049.81 ·
- Strata · IQ3_S (current)Decode 330.44 · Prefill 1792.04 ·
- SGLang · Q5_K_M Decode — · Prefill — ·
- SGLang · Q4_K_M Decode — · Prefill — ·
- vLLM · Q5_K_M Decode — · Prefill — ·
- vLLM · Q4_K_M Decode — · Prefill — ·
Core figures
- Decode (avg)
- 330.44 tok/s
- Prefill (avg)
- 1,792.04 tok/s
- TTFT (avg)
- 2.4 s
- VRAM (avg)
- 1.05 GB
- MTP acceptance rate
- 0.52
- TTFB
- 2.64 GB
- Power draw
- 460.5 W
Configuration
Member-level fields are taken from the most recent run
- Model
- Kimi-K3
- Quantization
- IQ3_S
- Framework
- Strata
- Version
- v0.39
- Context length
- 2048 tokens
- Batch size
- 1024
- GPU layers
- 15
- Flash Attention
- On
Hardware
Nominal and measured figures are shown side by side; whether it runs is the reader's call
- GPU
- NVIDIA RTX 5090
- Nominal VRAM
- 32 GB
- Measured VRAM (avg)
- 1.05 GB
- OS
- Ubuntu 22.04
- Driver
- 565.57
- CUDA
- 12.6
- Power draw
- 460.5 W
Sources & evidence
1 measured records in total, each traceable to its original source
v0.39 · Ubuntu 22.04 · CUDA 12.6 · 2048 ctx
330.44 tok/s
Decode
1,792.04 tok/s
Prefill
2.4 s
TTFT
1.05 GB
VRAM
0.52
MTP
460.5 W
Power draw
L2 batch seed 2026-10-18: Strata on hw_id=13 running model_id=73 at IQ3_S