llm-benchmarks
githubdrose.io

Kimi K2 Thinking Benchmarks


Measured on novita

4 runs
Avg Tokens / Second
5.02
Avg Time to First Token (ms)
11350.00
Runs Analysed
4
Last Updated
Sep 1, 2026, 03:11 AM

Throughput distribution

Throughput over time

Kimi K2 Thinkingnovita1tok/s

Generated throughput, including reasoning tokens — so a thinking model's line does not collapse; the table ranks on visible tokens · one cell per model, showing its best-covered provider · shared vertical scale, 0 to 39 tok/s (99th percentile) · dashed rule is the model's own mean · 1 further provider not drawn

Every provider serving this model

2 measured
ProviderModelAvg Toks/SecMinMaxAvg TTF (ms)
novitaKimi K2 Thinking4.734.005.2912200.00
novitaKimi K2 Thinking5.315.315.3110500.00

Frequently Asked Questions

How fast is Kimi K2 Thinking?
The latest rolling average throughput is 5.02 tokens per second with an average time to first token of 11350.00 ms across 4 recent runs.
How often are these benchmarks updated?
Benchmarks refresh automatically whenever the monitoring cron runs. The most recent run completed on Sep 1, 2026, 03:11 AM.

Related models