Tokens/Second Visualizer — Free & Secure | werkzeuge
Feel how fast an LLM responds at a given token rate. Streams text live in your browser — nothing is transmitted.
See and feel how fast a language model responds at a given token rate. The text streams live in your browser.
100% in your browser — nothing leaves your device.
Sample text
Long German compound words are split into many subword tokens — that is why they cost more than short English words.
Presets:
Adjustable during playback. Typical values: local/CPU ~10, cloud models ~50–120.
Total tokens
Shown
Elapsed (s)
Effective (tok/s)
Estimated duration (s)
Live stream
estimate
The token split is a heuristic and only an estimate — real tokenizers (OpenAI BPE, Claude, Gemini) count differently. For exact OpenAI token counts, use the Token Counter.
Go to Token Counter
AI Knowledge
Did you know?
Deep learning uses neural networks with many layers. The breakthrough came in 2012 when AlexNet won the ImageNet competition by a large margin.
Source: Krizhevsky, A. et al., NeurIPS, 2012
Workflow-Automatisierung
Connect AI models, data sources, and tools into powerful automated workflows.