llm-benchmarks
githubdrose.io

atlas-cloud Provider Benchmarks


Measured over the last 30 days

Models Tracked
24
Avg Tokens / Second
12.40
Avg Time to First Token (ms)
13407.52
Last Updated
Sep 2, 2026

All atlas-cloud models

24 tracked · fastest first
ModelAvg Toks/SecMinMaxAvg TTF (ms)
KAT-Coder-Pro V235.8031.6041.401220.00
DeepSeek V3.1 Terminus26.2017.3036.002380.00
Qwen3 235B A22B Instruct 250725.0022.1027.401280.00
DeepSeek V3.124.1014.6031.401960.00
GLM 5.222.8018.8025.801920.00
DeepSeek V3.2 Exp21.8016.3025.201850.00
DeepSeek V3.217.6013.7023.102020.00
DeepSeek V4 Flash 042316.9015.3018.303270.00
GLM 5.114.402.4723.208320.00
MiniMax M2.513.104.0928.408670.00
DeepSeek V4 Flash 073112.201.0325.5020100.00
MiMo-V2.5-Pro10.309.5411.204860.00
GLM 59.442.4014.1010900.00
Kimi K2.7 Code7.803.6310.009860.00
Kimi K2.55.644.247.0210500.00
MiniMax M2.75.553.829.1011700.00
LongCat 2.05.331.537.8216700.00
Qwen3.5-35B-A3B5.324.645.8711500.00
Qwen3.5-122B-A10B5.314.335.8712000.00
Kimi K2.64.452.436.9616000.00
GLM 4.74.073.285.5216100.00
MiniMax M33.922.905.9119100.00
Qwen3.5-27B3.723.334.1816800.00
Qwen3.5 397B A17B3.492.883.9818000.00

Frequently Asked Questions

Which atlas-cloud model is fastest?
Based on recent tests, KAT-Coder-Pro V2 shows the highest average throughput among tracked atlas-cloud models.
How many recent measurements feed this dashboard?
This provider summary aggregates 5434 individual prompts measured across 4510 monitoring runs over the past month.