llm-benchmarks
githubdrose.io

digitalocean Provider Benchmarks


Measured over the last 30 days

Models Tracked
26
Avg Tokens / Second
13.73
Avg Time to First Token (ms)
12909.45
Last Updated
Sep 1, 2026

All digitalocean models

26 tracked · fastest first
ModelAvg Toks/SecMinMaxAvg TTF (ms)
Qwen3 Coder 30B A3B Instruct73.9073.9073.90550.00
DeepSeek V4 Pro 042334.8028.4042.90650.00
DeepSeek V3.227.406.7745.102340.00
DeepSeek V4 Pro 042317.802.6744.901780.00
Qwen3.8 2.4T A95B17.809.7928.503470.00
MiMo-V2.5-Pro14.005.1717.204300.00
GLM 5.213.900.6033.0017000.00
Qwen3.8 2.4T A95B13.8010.7017.704090.00
GLM 5.211.302.2922.3010100.00
gpt-oss-120b11.107.3417.305210.00
Llama 4 Maverick9.856.3139.301160.00
Llama 4 Maverick9.773.7420.801210.00
DeepSeek V4 Flash 04236.892.2312.901370.00
DeepSeek V4 Flash 04236.122.249.642230.00
MiniMax M2.55.451.8110.3016100.00
Kimi K2.63.622.474.9117900.00
Kimi K2.52.471.125.6331400.00
DeepSeek V4 Flash 07312.360.927.9432400.00
DeepSeek V4 Pro 08132.312.312.3125800.00
DeepSeek V4 Flash 07312.271.702.7028100.00
Nemotron 3 Super1.871.142.5830900.00
Kimi K2.61.280.911.6451500.00
Kimi K2.51.211.211.2149600.00
Qwen3.5 397B A17B0.720.551.1388000.00
GLM 5.10.680.650.7188200.00
Qwen3.5 397B A17B0.660.630.6891700.00

Frequently Asked Questions

Which digitalocean model is fastest?
Based on recent tests, Qwen3 Coder 30B A3B Instruct shows the highest average throughput among tracked digitalocean models.
How many recent measurements feed this dashboard?
This provider summary aggregates 4754 individual prompts measured across 4002 monitoring runs over the past month.