llm-benchmarks
githubdrose.io

akashml Provider Benchmarks


Measured over the last 30 days

Models Tracked
15
Avg Tokens / Second
18.36
Avg Time to First Token (ms)
9981.75
Last Updated
Sep 1, 2026

All akashml models

15 tracked · fastest first
ModelAvg Toks/SecMinMaxAvg TTF (ms)
Llama 3.3 70B Instruct29.0019.0038.70960.00
gpt-oss-120b25.6025.6025.601060.00
GLM 5.225.4025.3025.401620.00
Llama 3.3 70B Instruct24.208.1634.401660.00
Llama 3.3 70B Instruct21.006.9334.601840.00
gpt-oss-120b16.106.8326.704100.00
gpt-oss-120b15.2015.2015.203200.00
Qwen3.8 27B12.1012.1012.104520.00
Qwen3.5-35B-A3B11.908.8714.505490.00
Qwen3.5-35B-A3B11.708.8115.305560.00
Qwen3.8 27B10.301.1614.9019800.00
Llama 3.3 70B Instruct8.908.908.905830.00
Qwen3.6 35B A3B7.656.5210.108350.00
Qwen3.6 35B A3B3.973.973.9716200.00
DeepSeek V4 Flash 07311.450.682.4552900.00

Frequently Asked Questions

Which akashml model is fastest?
Based on recent tests, Llama 3.3 70B Instruct shows the highest average throughput among tracked akashml models.
How many recent measurements feed this dashboard?
This provider summary aggregates 2231 individual prompts measured across 2083 monitoring runs over the past month.