LLM Speed Calculator
Turn a model's tokens-per-second rate into the real time it takes to generate a response.
This tokens per second calculator turns a model's streaming speed into a real wall-clock time. Enter the output tokens you expect, the decode rate in tokens per second, and the time to first token, and it estimates how long the full response takes to generate. It doubles as an llm response time estimator for chat interfaces, and it converts tokens per second to words per minute so you can picture the reading pace. Add your prompt length and a prefill rate to see the time to generate tokens end to end. Everything runs in your browser from the numbers you supply.
Read the guide: Tokens Per Second Explained (With a Wait-Time Example)Generation
Total time to generate
11 s
500 tokens at 50 tok/s
- Time to first token
- 0.4 s
- Prompt prefill
- 0.5 s
- Decode (streaming)
- 10 s
- Reading speed
- 2,250 wpm
Total
11 s
- First token4%
- Prefill5%
- Decode92%
TTFT is the wait before anything appears — queue time plus the model reading your prompt. Decode is the token-by-token streaming after that. For a short answer the first-token wait dominates; for a long one, decode speed is what you feel.
How it works
- 1
Set the response size and speed
Enter how many output tokens the reply will be and the decode rate in tokens per second. Drag the sliders for quick what-ifs or type exact figures from a benchmark.
- 2
Add latency and prompt
Set the time to first token in milliseconds, then your prompt length and a prefill rate. Prefill is usually far faster than decode; leave the input at zero to skip it.
- 3
Read the total time
The headline is the end-to-end wall-clock time. The stats split it into first token, prefill and decode, and show the reading speed in words per minute so you can feel how fast it streams.
Instant & 100% private — nothing is uploaded
Every calculation runs locally in your browser. The prompts, token counts and numbers you enter stay on your own device and are never sent to a server — nothing is stored, logged or shared.
Frequently asked questions
- How do you convert tokens per second to time?
- Decode time is output tokens divided by the tokens-per-second rate. Total time adds the time to first token and, if you include a prompt, the prefill time (input tokens divided by the prefill rate). So 500 tokens at 50 per second is 10 seconds of decode, plus whatever the first-token wait and prefill add.
- What is the difference between TTFT and decode?
- Time to first token is the wait before anything appears: queue time plus the model reading your prompt. Decode is the token-by-token streaming after that. For a short answer the first-token wait dominates the feel; for a long answer, decode speed is what you notice.
- How do tokens per second map to words per minute?
- Roughly 0.75 words per token, so a rate in tokens per second times 0.75 times 60 gives words per minute. Fifty tokens per second is about 2,250 words per minute, far faster than anyone reads, which is why streaming feels instant once it starts.
- Will the real time match this exactly?
- No. Actual speed swings with server load, batch size, prompt length, network latency and how the provider schedules your request. Use this as a planning estimate and measure your own numbers under real traffic before you rely on them.
Important
For planning and estimates only. Prices come from a published rate table dated on the page; providers change pricing without notice, and token counts here are approximations. Confirm against the provider’s own pricing before you budget or commit.
More tools
More from the Hivly network
Free sister tools on our other sites.
