llm-benchmarks
githubdrose.io

streamlake Provider Benchmarks


Measured over the last 30 days

Models Tracked
41
Avg Tokens / Second
13.18
Avg Time to First Token (ms)
12933.52
Last Updated
Sep 1, 2026

All streamlake models

41 tracked · fastest first
ModelAvg Toks/SecMinMaxAvg TTF (ms)
Qwen3 Coder Next38.5033.2042.601190.00
KAT-Coder-Pro V237.6027.5051.801180.00
KAT-Coder-Pro V237.4029.6042.101220.00
KAT-Coder-Air V2.536.8026.2042.001460.00
Qwen3 30B A3B Instruct 250735.4014.2059.001680.00
KAT-Coder-Air V2.534.0010.4046.501630.00
DeepSeek V3.1 Terminus29.3025.3033.001350.00
Qwen3 30B A3B Instruct 250724.801.9636.802070.00
Qwen3 235B A22B Instruct 250722.0019.6026.401400.00
DeepSeek V319.0015.0023.101690.00
GLM 5.217.4011.2022.102920.00
MiMo-V2.5-Pro16.9016.9016.902600.00
DeepSeek V3.216.8014.7018.401770.00
DeepSeek V316.507.4626.502680.00
DeepSeek V3.216.3016.3016.301880.00
GLM 5.215.602.2522.204480.00
MiniMax M315.208.9322.904700.00
MiMo-V2.5-Pro15.2013.5018.403030.00
DeepSeek V3.213.304.5118.304700.00
DeepSeek V4 Flash 073112.901.6425.206300.00
Qwen3.5 397B A17B11.6011.0012.405490.00
KAT-Coder-Pro V2.510.806.4513.406150.00
DeepSeek V4 Flash 07318.426.8410.806930.00
DeepSeek V4 Pro 04237.495.729.087080.00
MiniMax M2.56.103.1111.7013000.00
Kimi K2.56.064.498.809720.00
DeepSeek V4 Pro 04235.983.138.8312900.00
R1 05285.695.117.0710400.00
MiMo-V2.55.565.565.569630.00
KAT-Coder-Pro V2.55.075.075.0711900.00
DeepSeek V4 Flash 04234.594.075.3912400.00
Kimi K2.54.123.435.0213500.00
DeepSeek V4 Flash 04233.932.306.0315400.00
GLM 53.672.364.8917600.00
Kimi K2.63.612.206.5519300.00
GLM 53.443.084.1218100.00
Kimi K2.63.110.995.6221300.00
DeepSeek V4 Pro 08132.280.874.5841100.00
GLM 4.71.951.951.9531400.00
GLM 5.11.751.711.8034900.00
GLM 5.11.291.061.5348900.00

Frequently Asked Questions

Which streamlake model is fastest?
Based on recent tests, Qwen3 Coder Next shows the highest average throughput among tracked streamlake models.
How many recent measurements feed this dashboard?
This provider summary aggregates 7627 individual prompts measured across 5509 monitoring runs over the past month.