llm-benchmarks
githubdrose.io

parasail Provider Benchmarks


Measured over the last 30 days

Models Tracked
60
Avg Tokens / Second
18.55
Avg Time to First Token (ms)
9870.19
Last Updated
Sep 1, 2026

All parasail models

60 tracked · fastest first
ModelAvg Toks/SecMinMaxAvg TTF (ms)
Llama 3 8B Lunaris74.2074.2074.20580.00
Llama 3.2 3B Instruct59.0044.9077.40690.00
Qwen3 Coder Next50.3044.1056.30610.00
UnslopNemo 12B49.9038.1058.50680.00
Qwen3 Coder Next46.6028.0079.10820.00
UI-TARS 7B45.8037.8055.70670.00
MythoMax 13B44.2028.0059.60710.00
UnslopNemo 12B43.8013.5084.10920.00
Llama 3.2 3B Instruct43.5018.5087.301150.00
Qwen3 Next 80B A3B Instruct43.0032.2051.801060.00
Skyfall 36B V242.4033.3047.20630.00
Llama 3 8B Lunaris40.7021.5063.90860.00
Llama 4 Maverick38.2033.6048.00810.00
UI-TARS 7B35.7020.0082.201050.00
Mistral Nemo35.3022.9046.20670.00
Rocinante 12B35.0029.4040.00670.00
Rocinante 12B34.9013.0065.40690.00
MiniMax M332.6025.3039.201500.00
Llama 3.3 70B Instruct32.5020.6039.60850.00
Trinity Large Thinking31.3029.8032.901650.00
Qwen2.5 VL 72B Instruct30.509.3040.701290.00
Cydonia 24B V4.130.306.6746.201050.00
Trinity Large Thinking29.9025.9034.001770.00
Qwen2.5 VL 72B Instruct29.4015.8039.90630.00
Skyfall 36B V229.2010.4059.30900.00
Cydonia 24B V4.129.109.3244.401810.00
Mistral Small 3.2 24B28.7016.4042.001160.00
Qwen3 235B A22B Instruct 250728.2027.1029.50810.00
Gemma 3 27B26.7022.3030.80840.00
Gemma 4 26B A4B25.607.0346.00980.00
gpt-oss-120b25.2018.3032.802110.00
Qwen3 VL 235B A22B Instruct25.2011.6035.90730.00
Qwen3 VL 235B A22B Instruct24.9013.2031.601820.00
Gemma 3 27B24.005.1042.902710.00
Qwen3 VL 8B Instruct24.0012.1036.502100.00
gpt-oss-20b21.5021.5021.502450.00
Qwen3 235B A22B Instruct 250720.409.7526.102370.00
Gemma 4 31B18.208.7325.80960.00
Mistral Nemo17.706.7628.704860.00
GLM 5.217.402.7944.008780.00
DeepSeek V4 Pro 042312.3012.3012.304400.00
Kimi K2.612.106.9217.505020.00
MiMo-V2.512.107.2019.405350.00
Auto Router (Beta)11.709.7013.705230.00
gpt-oss-20b11.508.6613.904770.00
Muse Glimmer 30B10.6010.6010.605050.00
Kimi K2.7 Code10.402.9115.409460.00
Qwen3.5-35B-A3B10.109.4310.806060.00
DeepSeek V4 Flash 07319.433.9115.007400.00
DeepSeek V4 Flash 04235.924.648.9510300.00
Qwen3.5-35B-A3B5.174.017.3913100.00
Qwen3.6 35B A3B5.104.145.6012300.00
Qwen3.5-9B4.741.498.0720800.00
GLM 5.14.324.034.8213300.00
Qwen3.5 397B A17B4.242.766.1215500.00
DeepSeek V4 Pro 08133.551.885.0621000.00
Qwen3.8 27B3.493.493.4916900.00
Kimi K2.62.672.672.6723200.00
Qwen3.6 35B A3B2.332.332.3326600.00
Qwen3.5 397B A17B2.311.504.4631800.00

Frequently Asked Questions

Which parasail model is fastest?
Based on recent tests, Llama 3 8B Lunaris shows the highest average throughput among tracked parasail models.
How many recent measurements feed this dashboard?
This provider summary aggregates 6788 individual prompts measured across 5539 monitoring runs over the past month.