← All models

deepseek-v4-pro

1650.5B parameters FP8 1,048,576 token context completiontoolsthinking
107.4
output tokens / second
160 tokens · 1.49s server · 1.64s request

Performance statistics

Calculated live from valid twenty-minute RRDtool samples.

Range Latest Average Lowest Median P50 P05 Highest Std dev Variability CV Trend Data coverage
Last 24 hours 106.7 87.7 13.6 91.6 52.2 129.3 23.3 26.6% +13.1% 72 samples 100.0%
Last 7 days 106.7 88.7 4.6 94.7 31.4 137.0 27.0 30.5% -7.8% 504 samples 100.0%
Last 30 days 106.7 83.9 3.6 89.2 23.3 144.9 29.0 34.6% +3.6% 2160 samples 100.0%

P05: 5% of twenty-minute throughput samples are at or below this value; it represents lower-tail performance. CV: standard deviation ÷ average; lower is more stable. Trend: second-half average versus first-half average. Data coverage: valid samples versus all expected twenty-minute slots in the selected range.

Last 24 hours

deepseek-v4-pro throughput for Last 24 hours